xref: /openbmc/qemu/hw/arm/virt.c (revision 34aee9c9)
1 /*
2  * ARM mach-virt emulation
3  *
4  * Copyright (c) 2013 Linaro Limited
5  *
6  * This program is free software; you can redistribute it and/or modify it
7  * under the terms and conditions of the GNU General Public License,
8  * version 2 or later, as published by the Free Software Foundation.
9  *
10  * This program is distributed in the hope it will be useful, but WITHOUT
11  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
13  * more details.
14  *
15  * You should have received a copy of the GNU General Public License along with
16  * this program.  If not, see <http://www.gnu.org/licenses/>.
17  *
18  * Emulate a virtual board which works by passing Linux all the information
19  * it needs about what devices are present via the device tree.
20  * There are some restrictions about what we can do here:
21  *  + we can only present devices whose Linux drivers will work based
22  *    purely on the device tree with no platform data at all
23  *  + we want to present a very stripped-down minimalist platform,
24  *    both because this reduces the security attack surface from the guest
25  *    and also because it reduces our exposure to being broken when
26  *    the kernel updates its device tree bindings and requires further
27  *    information in a device binding that we aren't providing.
28  * This is essentially the same approach kvmtool uses.
29  */
30 
31 #include "qemu/osdep.h"
32 #include "qemu/datadir.h"
33 #include "qemu/units.h"
34 #include "qemu/option.h"
35 #include "monitor/qdev.h"
36 #include "hw/sysbus.h"
37 #include "hw/arm/boot.h"
38 #include "hw/arm/primecell.h"
39 #include "hw/arm/virt.h"
40 #include "hw/block/flash.h"
41 #include "hw/vfio/vfio-calxeda-xgmac.h"
42 #include "hw/vfio/vfio-amd-xgbe.h"
43 #include "hw/display/ramfb.h"
44 #include "net/net.h"
45 #include "sysemu/device_tree.h"
46 #include "sysemu/numa.h"
47 #include "sysemu/runstate.h"
48 #include "sysemu/tpm.h"
49 #include "sysemu/tcg.h"
50 #include "sysemu/kvm.h"
51 #include "sysemu/hvf.h"
52 #include "sysemu/qtest.h"
53 #include "hw/loader.h"
54 #include "qapi/error.h"
55 #include "qemu/bitops.h"
56 #include "qemu/error-report.h"
57 #include "qemu/module.h"
58 #include "hw/pci-host/gpex.h"
59 #include "hw/virtio/virtio-pci.h"
60 #include "hw/core/sysbus-fdt.h"
61 #include "hw/platform-bus.h"
62 #include "hw/qdev-properties.h"
63 #include "hw/arm/fdt.h"
64 #include "hw/intc/arm_gic.h"
65 #include "hw/intc/arm_gicv3_common.h"
66 #include "hw/intc/arm_gicv3_its_common.h"
67 #include "hw/irq.h"
68 #include "kvm_arm.h"
69 #include "hw/firmware/smbios.h"
70 #include "qapi/visitor.h"
71 #include "qapi/qapi-visit-common.h"
72 #include "qapi/qmp/qlist.h"
73 #include "standard-headers/linux/input.h"
74 #include "hw/arm/smmuv3.h"
75 #include "hw/acpi/acpi.h"
76 #include "target/arm/internals.h"
77 #include "hw/mem/pc-dimm.h"
78 #include "hw/mem/nvdimm.h"
79 #include "hw/acpi/generic_event_device.h"
80 #include "hw/virtio/virtio-md-pci.h"
81 #include "hw/virtio/virtio-iommu.h"
82 #include "hw/char/pl011.h"
83 #include "qemu/guest-random.h"
84 
85 #define DEFINE_VIRT_MACHINE_LATEST(major, minor, latest) \
86     static void virt_##major##_##minor##_class_init(ObjectClass *oc, \
87                                                     void *data) \
88     { \
89         MachineClass *mc = MACHINE_CLASS(oc); \
90         virt_machine_##major##_##minor##_options(mc); \
91         mc->desc = "QEMU " # major "." # minor " ARM Virtual Machine"; \
92         if (latest) { \
93             mc->alias = "virt"; \
94         } \
95     } \
96     static const TypeInfo machvirt_##major##_##minor##_info = { \
97         .name = MACHINE_TYPE_NAME("virt-" # major "." # minor), \
98         .parent = TYPE_VIRT_MACHINE, \
99         .class_init = virt_##major##_##minor##_class_init, \
100     }; \
101     static void machvirt_machine_##major##_##minor##_init(void) \
102     { \
103         type_register_static(&machvirt_##major##_##minor##_info); \
104     } \
105     type_init(machvirt_machine_##major##_##minor##_init);
106 
107 #define DEFINE_VIRT_MACHINE_AS_LATEST(major, minor) \
108     DEFINE_VIRT_MACHINE_LATEST(major, minor, true)
109 #define DEFINE_VIRT_MACHINE(major, minor) \
110     DEFINE_VIRT_MACHINE_LATEST(major, minor, false)
111 
112 
113 /* Number of external interrupt lines to configure the GIC with */
114 #define NUM_IRQS 256
115 
116 #define PLATFORM_BUS_NUM_IRQS 64
117 
118 /* Legacy RAM limit in GB (< version 4.0) */
119 #define LEGACY_RAMLIMIT_GB 255
120 #define LEGACY_RAMLIMIT_BYTES (LEGACY_RAMLIMIT_GB * GiB)
121 
122 /* Addresses and sizes of our components.
123  * 0..128MB is space for a flash device so we can run bootrom code such as UEFI.
124  * 128MB..256MB is used for miscellaneous device I/O.
125  * 256MB..1GB is reserved for possible future PCI support (ie where the
126  * PCI memory window will go if we add a PCI host controller).
127  * 1GB and up is RAM (which may happily spill over into the
128  * high memory region beyond 4GB).
129  * This represents a compromise between how much RAM can be given to
130  * a 32 bit VM and leaving space for expansion and in particular for PCI.
131  * Note that devices should generally be placed at multiples of 0x10000,
132  * to accommodate guests using 64K pages.
133  */
134 static const MemMapEntry base_memmap[] = {
135     /* Space up to 0x8000000 is reserved for a boot ROM */
136     [VIRT_FLASH] =              {          0, 0x08000000 },
137     [VIRT_CPUPERIPHS] =         { 0x08000000, 0x00020000 },
138     /* GIC distributor and CPU interfaces sit inside the CPU peripheral space */
139     [VIRT_GIC_DIST] =           { 0x08000000, 0x00010000 },
140     [VIRT_GIC_CPU] =            { 0x08010000, 0x00010000 },
141     [VIRT_GIC_V2M] =            { 0x08020000, 0x00001000 },
142     [VIRT_GIC_HYP] =            { 0x08030000, 0x00010000 },
143     [VIRT_GIC_VCPU] =           { 0x08040000, 0x00010000 },
144     /* The space in between here is reserved for GICv3 CPU/vCPU/HYP */
145     [VIRT_GIC_ITS] =            { 0x08080000, 0x00020000 },
146     /* This redistributor space allows up to 2*64kB*123 CPUs */
147     [VIRT_GIC_REDIST] =         { 0x080A0000, 0x00F60000 },
148     [VIRT_UART] =               { 0x09000000, 0x00001000 },
149     [VIRT_RTC] =                { 0x09010000, 0x00001000 },
150     [VIRT_FW_CFG] =             { 0x09020000, 0x00000018 },
151     [VIRT_GPIO] =               { 0x09030000, 0x00001000 },
152     [VIRT_SECURE_UART] =        { 0x09040000, 0x00001000 },
153     [VIRT_SMMU] =               { 0x09050000, 0x00020000 },
154     [VIRT_PCDIMM_ACPI] =        { 0x09070000, MEMORY_HOTPLUG_IO_LEN },
155     [VIRT_ACPI_GED] =           { 0x09080000, ACPI_GED_EVT_SEL_LEN },
156     [VIRT_NVDIMM_ACPI] =        { 0x09090000, NVDIMM_ACPI_IO_LEN},
157     [VIRT_PVTIME] =             { 0x090a0000, 0x00010000 },
158     [VIRT_SECURE_GPIO] =        { 0x090b0000, 0x00001000 },
159     [VIRT_MMIO] =               { 0x0a000000, 0x00000200 },
160     /* ...repeating for a total of NUM_VIRTIO_TRANSPORTS, each of that size */
161     [VIRT_PLATFORM_BUS] =       { 0x0c000000, 0x02000000 },
162     [VIRT_SECURE_MEM] =         { 0x0e000000, 0x01000000 },
163     [VIRT_PCIE_MMIO] =          { 0x10000000, 0x2eff0000 },
164     [VIRT_PCIE_PIO] =           { 0x3eff0000, 0x00010000 },
165     [VIRT_PCIE_ECAM] =          { 0x3f000000, 0x01000000 },
166     /* Actual RAM size depends on initial RAM and device memory settings */
167     [VIRT_MEM] =                { GiB, LEGACY_RAMLIMIT_BYTES },
168 };
169 
170 /*
171  * Highmem IO Regions: This memory map is floating, located after the RAM.
172  * Each MemMapEntry base (GPA) will be dynamically computed, depending on the
173  * top of the RAM, so that its base get the same alignment as the size,
174  * ie. a 512GiB entry will be aligned on a 512GiB boundary. If there is
175  * less than 256GiB of RAM, the floating area starts at the 256GiB mark.
176  * Note the extended_memmap is sized so that it eventually also includes the
177  * base_memmap entries (VIRT_HIGH_GIC_REDIST2 index is greater than the last
178  * index of base_memmap).
179  *
180  * The memory map for these Highmem IO Regions can be in legacy or compact
181  * layout, depending on 'compact-highmem' property. With legacy layout, the
182  * PA space for one specific region is always reserved, even if the region
183  * has been disabled or doesn't fit into the PA space. However, the PA space
184  * for the region won't be reserved in these circumstances with compact layout.
185  */
186 static MemMapEntry extended_memmap[] = {
187     /* Additional 64 MB redist region (can contain up to 512 redistributors) */
188     [VIRT_HIGH_GIC_REDIST2] =   { 0x0, 64 * MiB },
189     [VIRT_HIGH_PCIE_ECAM] =     { 0x0, 256 * MiB },
190     /* Second PCIe window */
191     [VIRT_HIGH_PCIE_MMIO] =     { 0x0, 512 * GiB },
192 };
193 
194 static const int a15irqmap[] = {
195     [VIRT_UART] = 1,
196     [VIRT_RTC] = 2,
197     [VIRT_PCIE] = 3, /* ... to 6 */
198     [VIRT_GPIO] = 7,
199     [VIRT_SECURE_UART] = 8,
200     [VIRT_ACPI_GED] = 9,
201     [VIRT_MMIO] = 16, /* ...to 16 + NUM_VIRTIO_TRANSPORTS - 1 */
202     [VIRT_GIC_V2M] = 48, /* ...to 48 + NUM_GICV2M_SPIS - 1 */
203     [VIRT_SMMU] = 74,    /* ...to 74 + NUM_SMMU_IRQS - 1 */
204     [VIRT_PLATFORM_BUS] = 112, /* ...to 112 + PLATFORM_BUS_NUM_IRQS -1 */
205 };
206 
207 static const char *valid_cpus[] = {
208 #ifdef CONFIG_TCG
209     ARM_CPU_TYPE_NAME("cortex-a7"),
210     ARM_CPU_TYPE_NAME("cortex-a15"),
211     ARM_CPU_TYPE_NAME("cortex-a35"),
212     ARM_CPU_TYPE_NAME("cortex-a55"),
213     ARM_CPU_TYPE_NAME("cortex-a72"),
214     ARM_CPU_TYPE_NAME("cortex-a76"),
215     ARM_CPU_TYPE_NAME("cortex-a710"),
216     ARM_CPU_TYPE_NAME("a64fx"),
217     ARM_CPU_TYPE_NAME("neoverse-n1"),
218     ARM_CPU_TYPE_NAME("neoverse-v1"),
219     ARM_CPU_TYPE_NAME("neoverse-n2"),
220 #endif
221     ARM_CPU_TYPE_NAME("cortex-a53"),
222     ARM_CPU_TYPE_NAME("cortex-a57"),
223     ARM_CPU_TYPE_NAME("host"),
224     ARM_CPU_TYPE_NAME("max"),
225 };
226 
227 static bool cpu_type_valid(const char *cpu)
228 {
229     int i;
230 
231     for (i = 0; i < ARRAY_SIZE(valid_cpus); i++) {
232         if (strcmp(cpu, valid_cpus[i]) == 0) {
233             return true;
234         }
235     }
236     return false;
237 }
238 
239 static void create_randomness(MachineState *ms, const char *node)
240 {
241     struct {
242         uint64_t kaslr;
243         uint8_t rng[32];
244     } seed;
245 
246     if (qemu_guest_getrandom(&seed, sizeof(seed), NULL)) {
247         return;
248     }
249     qemu_fdt_setprop_u64(ms->fdt, node, "kaslr-seed", seed.kaslr);
250     qemu_fdt_setprop(ms->fdt, node, "rng-seed", seed.rng, sizeof(seed.rng));
251 }
252 
253 static void create_fdt(VirtMachineState *vms)
254 {
255     MachineState *ms = MACHINE(vms);
256     int nb_numa_nodes = ms->numa_state->num_nodes;
257     void *fdt = create_device_tree(&vms->fdt_size);
258 
259     if (!fdt) {
260         error_report("create_device_tree() failed");
261         exit(1);
262     }
263 
264     ms->fdt = fdt;
265 
266     /* Header */
267     qemu_fdt_setprop_string(fdt, "/", "compatible", "linux,dummy-virt");
268     qemu_fdt_setprop_cell(fdt, "/", "#address-cells", 0x2);
269     qemu_fdt_setprop_cell(fdt, "/", "#size-cells", 0x2);
270     qemu_fdt_setprop_string(fdt, "/", "model", "linux,dummy-virt");
271 
272     /* /chosen must exist for load_dtb to fill in necessary properties later */
273     qemu_fdt_add_subnode(fdt, "/chosen");
274     if (vms->dtb_randomness) {
275         create_randomness(ms, "/chosen");
276     }
277 
278     if (vms->secure) {
279         qemu_fdt_add_subnode(fdt, "/secure-chosen");
280         if (vms->dtb_randomness) {
281             create_randomness(ms, "/secure-chosen");
282         }
283     }
284 
285     /* Clock node, for the benefit of the UART. The kernel device tree
286      * binding documentation claims the PL011 node clock properties are
287      * optional but in practice if you omit them the kernel refuses to
288      * probe for the device.
289      */
290     vms->clock_phandle = qemu_fdt_alloc_phandle(fdt);
291     qemu_fdt_add_subnode(fdt, "/apb-pclk");
292     qemu_fdt_setprop_string(fdt, "/apb-pclk", "compatible", "fixed-clock");
293     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "#clock-cells", 0x0);
294     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "clock-frequency", 24000000);
295     qemu_fdt_setprop_string(fdt, "/apb-pclk", "clock-output-names",
296                                 "clk24mhz");
297     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "phandle", vms->clock_phandle);
298 
299     if (nb_numa_nodes > 0 && ms->numa_state->have_numa_distance) {
300         int size = nb_numa_nodes * nb_numa_nodes * 3 * sizeof(uint32_t);
301         uint32_t *matrix = g_malloc0(size);
302         int idx, i, j;
303 
304         for (i = 0; i < nb_numa_nodes; i++) {
305             for (j = 0; j < nb_numa_nodes; j++) {
306                 idx = (i * nb_numa_nodes + j) * 3;
307                 matrix[idx + 0] = cpu_to_be32(i);
308                 matrix[idx + 1] = cpu_to_be32(j);
309                 matrix[idx + 2] =
310                     cpu_to_be32(ms->numa_state->nodes[i].distance[j]);
311             }
312         }
313 
314         qemu_fdt_add_subnode(fdt, "/distance-map");
315         qemu_fdt_setprop_string(fdt, "/distance-map", "compatible",
316                                 "numa-distance-map-v1");
317         qemu_fdt_setprop(fdt, "/distance-map", "distance-matrix",
318                          matrix, size);
319         g_free(matrix);
320     }
321 }
322 
323 static void fdt_add_timer_nodes(const VirtMachineState *vms)
324 {
325     /* On real hardware these interrupts are level-triggered.
326      * On KVM they were edge-triggered before host kernel version 4.4,
327      * and level-triggered afterwards.
328      * On emulated QEMU they are level-triggered.
329      *
330      * Getting the DTB info about them wrong is awkward for some
331      * guest kernels:
332      *  pre-4.8 ignore the DT and leave the interrupt configured
333      *   with whatever the GIC reset value (or the bootloader) left it at
334      *  4.8 before rc6 honour the incorrect data by programming it back
335      *   into the GIC, causing problems
336      *  4.8rc6 and later ignore the DT and always write "level triggered"
337      *   into the GIC
338      *
339      * For backwards-compatibility, virt-2.8 and earlier will continue
340      * to say these are edge-triggered, but later machines will report
341      * the correct information.
342      */
343     ARMCPU *armcpu;
344     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
345     uint32_t irqflags = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
346     MachineState *ms = MACHINE(vms);
347 
348     if (vmc->claim_edge_triggered_timers) {
349         irqflags = GIC_FDT_IRQ_FLAGS_EDGE_LO_HI;
350     }
351 
352     if (vms->gic_version == VIRT_GIC_VERSION_2) {
353         irqflags = deposit32(irqflags, GIC_FDT_IRQ_PPI_CPU_START,
354                              GIC_FDT_IRQ_PPI_CPU_WIDTH,
355                              (1 << MACHINE(vms)->smp.cpus) - 1);
356     }
357 
358     qemu_fdt_add_subnode(ms->fdt, "/timer");
359 
360     armcpu = ARM_CPU(qemu_get_cpu(0));
361     if (arm_feature(&armcpu->env, ARM_FEATURE_V8)) {
362         const char compat[] = "arm,armv8-timer\0arm,armv7-timer";
363         qemu_fdt_setprop(ms->fdt, "/timer", "compatible",
364                          compat, sizeof(compat));
365     } else {
366         qemu_fdt_setprop_string(ms->fdt, "/timer", "compatible",
367                                 "arm,armv7-timer");
368     }
369     qemu_fdt_setprop(ms->fdt, "/timer", "always-on", NULL, 0);
370     qemu_fdt_setprop_cells(ms->fdt, "/timer", "interrupts",
371                            GIC_FDT_IRQ_TYPE_PPI,
372                            INTID_TO_PPI(ARCH_TIMER_S_EL1_IRQ), irqflags,
373                            GIC_FDT_IRQ_TYPE_PPI,
374                            INTID_TO_PPI(ARCH_TIMER_NS_EL1_IRQ), irqflags,
375                            GIC_FDT_IRQ_TYPE_PPI,
376                            INTID_TO_PPI(ARCH_TIMER_VIRT_IRQ), irqflags,
377                            GIC_FDT_IRQ_TYPE_PPI,
378                            INTID_TO_PPI(ARCH_TIMER_NS_EL2_IRQ), irqflags);
379 }
380 
381 static void fdt_add_cpu_nodes(const VirtMachineState *vms)
382 {
383     int cpu;
384     int addr_cells = 1;
385     const MachineState *ms = MACHINE(vms);
386     const VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
387     int smp_cpus = ms->smp.cpus;
388 
389     /*
390      * See Linux Documentation/devicetree/bindings/arm/cpus.yaml
391      * On ARM v8 64-bit systems value should be set to 2,
392      * that corresponds to the MPIDR_EL1 register size.
393      * If MPIDR_EL1[63:32] value is equal to 0 on all CPUs
394      * in the system, #address-cells can be set to 1, since
395      * MPIDR_EL1[63:32] bits are not used for CPUs
396      * identification.
397      *
398      * Here we actually don't know whether our system is 32- or 64-bit one.
399      * The simplest way to go is to examine affinity IDs of all our CPUs. If
400      * at least one of them has Aff3 populated, we set #address-cells to 2.
401      */
402     for (cpu = 0; cpu < smp_cpus; cpu++) {
403         ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
404 
405         if (armcpu->mp_affinity & ARM_AFF3_MASK) {
406             addr_cells = 2;
407             break;
408         }
409     }
410 
411     qemu_fdt_add_subnode(ms->fdt, "/cpus");
412     qemu_fdt_setprop_cell(ms->fdt, "/cpus", "#address-cells", addr_cells);
413     qemu_fdt_setprop_cell(ms->fdt, "/cpus", "#size-cells", 0x0);
414 
415     for (cpu = smp_cpus - 1; cpu >= 0; cpu--) {
416         char *nodename = g_strdup_printf("/cpus/cpu@%d", cpu);
417         ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
418         CPUState *cs = CPU(armcpu);
419 
420         qemu_fdt_add_subnode(ms->fdt, nodename);
421         qemu_fdt_setprop_string(ms->fdt, nodename, "device_type", "cpu");
422         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
423                                     armcpu->dtb_compatible);
424 
425         if (vms->psci_conduit != QEMU_PSCI_CONDUIT_DISABLED && smp_cpus > 1) {
426             qemu_fdt_setprop_string(ms->fdt, nodename,
427                                         "enable-method", "psci");
428         }
429 
430         if (addr_cells == 2) {
431             qemu_fdt_setprop_u64(ms->fdt, nodename, "reg",
432                                  armcpu->mp_affinity);
433         } else {
434             qemu_fdt_setprop_cell(ms->fdt, nodename, "reg",
435                                   armcpu->mp_affinity);
436         }
437 
438         if (ms->possible_cpus->cpus[cs->cpu_index].props.has_node_id) {
439             qemu_fdt_setprop_cell(ms->fdt, nodename, "numa-node-id",
440                 ms->possible_cpus->cpus[cs->cpu_index].props.node_id);
441         }
442 
443         if (!vmc->no_cpu_topology) {
444             qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle",
445                                   qemu_fdt_alloc_phandle(ms->fdt));
446         }
447 
448         g_free(nodename);
449     }
450 
451     if (!vmc->no_cpu_topology) {
452         /*
453          * Add vCPU topology description through fdt node cpu-map.
454          *
455          * See Linux Documentation/devicetree/bindings/cpu/cpu-topology.txt
456          * In a SMP system, the hierarchy of CPUs can be defined through
457          * four entities that are used to describe the layout of CPUs in
458          * the system: socket/cluster/core/thread.
459          *
460          * A socket node represents the boundary of system physical package
461          * and its child nodes must be one or more cluster nodes. A system
462          * can contain several layers of clustering within a single physical
463          * package and cluster nodes can be contained in parent cluster nodes.
464          *
465          * Note: currently we only support one layer of clustering within
466          * each physical package.
467          */
468         qemu_fdt_add_subnode(ms->fdt, "/cpus/cpu-map");
469 
470         for (cpu = smp_cpus - 1; cpu >= 0; cpu--) {
471             char *cpu_path = g_strdup_printf("/cpus/cpu@%d", cpu);
472             char *map_path;
473 
474             if (ms->smp.threads > 1) {
475                 map_path = g_strdup_printf(
476                     "/cpus/cpu-map/socket%d/cluster%d/core%d/thread%d",
477                     cpu / (ms->smp.clusters * ms->smp.cores * ms->smp.threads),
478                     (cpu / (ms->smp.cores * ms->smp.threads)) % ms->smp.clusters,
479                     (cpu / ms->smp.threads) % ms->smp.cores,
480                     cpu % ms->smp.threads);
481             } else {
482                 map_path = g_strdup_printf(
483                     "/cpus/cpu-map/socket%d/cluster%d/core%d",
484                     cpu / (ms->smp.clusters * ms->smp.cores),
485                     (cpu / ms->smp.cores) % ms->smp.clusters,
486                     cpu % ms->smp.cores);
487             }
488             qemu_fdt_add_path(ms->fdt, map_path);
489             qemu_fdt_setprop_phandle(ms->fdt, map_path, "cpu", cpu_path);
490 
491             g_free(map_path);
492             g_free(cpu_path);
493         }
494     }
495 }
496 
497 static void fdt_add_its_gic_node(VirtMachineState *vms)
498 {
499     char *nodename;
500     MachineState *ms = MACHINE(vms);
501 
502     vms->msi_phandle = qemu_fdt_alloc_phandle(ms->fdt);
503     nodename = g_strdup_printf("/intc/its@%" PRIx64,
504                                vms->memmap[VIRT_GIC_ITS].base);
505     qemu_fdt_add_subnode(ms->fdt, nodename);
506     qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
507                             "arm,gic-v3-its");
508     qemu_fdt_setprop(ms->fdt, nodename, "msi-controller", NULL, 0);
509     qemu_fdt_setprop_cell(ms->fdt, nodename, "#msi-cells", 1);
510     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
511                                  2, vms->memmap[VIRT_GIC_ITS].base,
512                                  2, vms->memmap[VIRT_GIC_ITS].size);
513     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", vms->msi_phandle);
514     g_free(nodename);
515 }
516 
517 static void fdt_add_v2m_gic_node(VirtMachineState *vms)
518 {
519     MachineState *ms = MACHINE(vms);
520     char *nodename;
521 
522     nodename = g_strdup_printf("/intc/v2m@%" PRIx64,
523                                vms->memmap[VIRT_GIC_V2M].base);
524     vms->msi_phandle = qemu_fdt_alloc_phandle(ms->fdt);
525     qemu_fdt_add_subnode(ms->fdt, nodename);
526     qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
527                             "arm,gic-v2m-frame");
528     qemu_fdt_setprop(ms->fdt, nodename, "msi-controller", NULL, 0);
529     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
530                                  2, vms->memmap[VIRT_GIC_V2M].base,
531                                  2, vms->memmap[VIRT_GIC_V2M].size);
532     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", vms->msi_phandle);
533     g_free(nodename);
534 }
535 
536 static void fdt_add_gic_node(VirtMachineState *vms)
537 {
538     MachineState *ms = MACHINE(vms);
539     char *nodename;
540 
541     vms->gic_phandle = qemu_fdt_alloc_phandle(ms->fdt);
542     qemu_fdt_setprop_cell(ms->fdt, "/", "interrupt-parent", vms->gic_phandle);
543 
544     nodename = g_strdup_printf("/intc@%" PRIx64,
545                                vms->memmap[VIRT_GIC_DIST].base);
546     qemu_fdt_add_subnode(ms->fdt, nodename);
547     qemu_fdt_setprop_cell(ms->fdt, nodename, "#interrupt-cells", 3);
548     qemu_fdt_setprop(ms->fdt, nodename, "interrupt-controller", NULL, 0);
549     qemu_fdt_setprop_cell(ms->fdt, nodename, "#address-cells", 0x2);
550     qemu_fdt_setprop_cell(ms->fdt, nodename, "#size-cells", 0x2);
551     qemu_fdt_setprop(ms->fdt, nodename, "ranges", NULL, 0);
552     if (vms->gic_version != VIRT_GIC_VERSION_2) {
553         int nb_redist_regions = virt_gicv3_redist_region_count(vms);
554 
555         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
556                                 "arm,gic-v3");
557 
558         qemu_fdt_setprop_cell(ms->fdt, nodename,
559                               "#redistributor-regions", nb_redist_regions);
560 
561         if (nb_redist_regions == 1) {
562             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
563                                          2, vms->memmap[VIRT_GIC_DIST].base,
564                                          2, vms->memmap[VIRT_GIC_DIST].size,
565                                          2, vms->memmap[VIRT_GIC_REDIST].base,
566                                          2, vms->memmap[VIRT_GIC_REDIST].size);
567         } else {
568             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
569                                  2, vms->memmap[VIRT_GIC_DIST].base,
570                                  2, vms->memmap[VIRT_GIC_DIST].size,
571                                  2, vms->memmap[VIRT_GIC_REDIST].base,
572                                  2, vms->memmap[VIRT_GIC_REDIST].size,
573                                  2, vms->memmap[VIRT_HIGH_GIC_REDIST2].base,
574                                  2, vms->memmap[VIRT_HIGH_GIC_REDIST2].size);
575         }
576 
577         if (vms->virt) {
578             qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
579                                    GIC_FDT_IRQ_TYPE_PPI, ARCH_GIC_MAINT_IRQ,
580                                    GIC_FDT_IRQ_FLAGS_LEVEL_HI);
581         }
582     } else {
583         /* 'cortex-a15-gic' means 'GIC v2' */
584         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
585                                 "arm,cortex-a15-gic");
586         if (!vms->virt) {
587             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
588                                          2, vms->memmap[VIRT_GIC_DIST].base,
589                                          2, vms->memmap[VIRT_GIC_DIST].size,
590                                          2, vms->memmap[VIRT_GIC_CPU].base,
591                                          2, vms->memmap[VIRT_GIC_CPU].size);
592         } else {
593             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
594                                          2, vms->memmap[VIRT_GIC_DIST].base,
595                                          2, vms->memmap[VIRT_GIC_DIST].size,
596                                          2, vms->memmap[VIRT_GIC_CPU].base,
597                                          2, vms->memmap[VIRT_GIC_CPU].size,
598                                          2, vms->memmap[VIRT_GIC_HYP].base,
599                                          2, vms->memmap[VIRT_GIC_HYP].size,
600                                          2, vms->memmap[VIRT_GIC_VCPU].base,
601                                          2, vms->memmap[VIRT_GIC_VCPU].size);
602             qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
603                                    GIC_FDT_IRQ_TYPE_PPI, ARCH_GIC_MAINT_IRQ,
604                                    GIC_FDT_IRQ_FLAGS_LEVEL_HI);
605         }
606     }
607 
608     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", vms->gic_phandle);
609     g_free(nodename);
610 }
611 
612 static void fdt_add_pmu_nodes(const VirtMachineState *vms)
613 {
614     ARMCPU *armcpu = ARM_CPU(first_cpu);
615     uint32_t irqflags = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
616     MachineState *ms = MACHINE(vms);
617 
618     if (!arm_feature(&armcpu->env, ARM_FEATURE_PMU)) {
619         assert(!object_property_get_bool(OBJECT(armcpu), "pmu", NULL));
620         return;
621     }
622 
623     if (vms->gic_version == VIRT_GIC_VERSION_2) {
624         irqflags = deposit32(irqflags, GIC_FDT_IRQ_PPI_CPU_START,
625                              GIC_FDT_IRQ_PPI_CPU_WIDTH,
626                              (1 << MACHINE(vms)->smp.cpus) - 1);
627     }
628 
629     qemu_fdt_add_subnode(ms->fdt, "/pmu");
630     if (arm_feature(&armcpu->env, ARM_FEATURE_V8)) {
631         const char compat[] = "arm,armv8-pmuv3";
632         qemu_fdt_setprop(ms->fdt, "/pmu", "compatible",
633                          compat, sizeof(compat));
634         qemu_fdt_setprop_cells(ms->fdt, "/pmu", "interrupts",
635                                GIC_FDT_IRQ_TYPE_PPI,
636                                INTID_TO_PPI(VIRTUAL_PMU_IRQ), irqflags);
637     }
638 }
639 
640 static inline DeviceState *create_acpi_ged(VirtMachineState *vms)
641 {
642     DeviceState *dev;
643     MachineState *ms = MACHINE(vms);
644     int irq = vms->irqmap[VIRT_ACPI_GED];
645     uint32_t event = ACPI_GED_PWR_DOWN_EVT;
646 
647     if (ms->ram_slots) {
648         event |= ACPI_GED_MEM_HOTPLUG_EVT;
649     }
650 
651     if (ms->nvdimms_state->is_enabled) {
652         event |= ACPI_GED_NVDIMM_HOTPLUG_EVT;
653     }
654 
655     dev = qdev_new(TYPE_ACPI_GED);
656     qdev_prop_set_uint32(dev, "ged-event", event);
657     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
658 
659     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_ACPI_GED].base);
660     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 1, vms->memmap[VIRT_PCDIMM_ACPI].base);
661     sysbus_connect_irq(SYS_BUS_DEVICE(dev), 0, qdev_get_gpio_in(vms->gic, irq));
662 
663     return dev;
664 }
665 
666 static void create_its(VirtMachineState *vms)
667 {
668     const char *itsclass = its_class_name();
669     DeviceState *dev;
670 
671     if (!strcmp(itsclass, "arm-gicv3-its")) {
672         if (!vms->tcg_its) {
673             itsclass = NULL;
674         }
675     }
676 
677     if (!itsclass) {
678         /* Do nothing if not supported */
679         return;
680     }
681 
682     dev = qdev_new(itsclass);
683 
684     object_property_set_link(OBJECT(dev), "parent-gicv3", OBJECT(vms->gic),
685                              &error_abort);
686     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
687     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_ITS].base);
688 
689     fdt_add_its_gic_node(vms);
690     vms->msi_controller = VIRT_MSI_CTRL_ITS;
691 }
692 
693 static void create_v2m(VirtMachineState *vms)
694 {
695     int i;
696     int irq = vms->irqmap[VIRT_GIC_V2M];
697     DeviceState *dev;
698 
699     dev = qdev_new("arm-gicv2m");
700     qdev_prop_set_uint32(dev, "base-spi", irq);
701     qdev_prop_set_uint32(dev, "num-spi", NUM_GICV2M_SPIS);
702     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
703     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_V2M].base);
704 
705     for (i = 0; i < NUM_GICV2M_SPIS; i++) {
706         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
707                            qdev_get_gpio_in(vms->gic, irq + i));
708     }
709 
710     fdt_add_v2m_gic_node(vms);
711     vms->msi_controller = VIRT_MSI_CTRL_GICV2M;
712 }
713 
714 static void create_gic(VirtMachineState *vms, MemoryRegion *mem)
715 {
716     MachineState *ms = MACHINE(vms);
717     /* We create a standalone GIC */
718     SysBusDevice *gicbusdev;
719     const char *gictype;
720     int i;
721     unsigned int smp_cpus = ms->smp.cpus;
722     uint32_t nb_redist_regions = 0;
723     int revision;
724 
725     if (vms->gic_version == VIRT_GIC_VERSION_2) {
726         gictype = gic_class_name();
727     } else {
728         gictype = gicv3_class_name();
729     }
730 
731     switch (vms->gic_version) {
732     case VIRT_GIC_VERSION_2:
733         revision = 2;
734         break;
735     case VIRT_GIC_VERSION_3:
736         revision = 3;
737         break;
738     case VIRT_GIC_VERSION_4:
739         revision = 4;
740         break;
741     default:
742         g_assert_not_reached();
743     }
744     vms->gic = qdev_new(gictype);
745     qdev_prop_set_uint32(vms->gic, "revision", revision);
746     qdev_prop_set_uint32(vms->gic, "num-cpu", smp_cpus);
747     /* Note that the num-irq property counts both internal and external
748      * interrupts; there are always 32 of the former (mandated by GIC spec).
749      */
750     qdev_prop_set_uint32(vms->gic, "num-irq", NUM_IRQS + 32);
751     if (!kvm_irqchip_in_kernel()) {
752         qdev_prop_set_bit(vms->gic, "has-security-extensions", vms->secure);
753     }
754 
755     if (vms->gic_version != VIRT_GIC_VERSION_2) {
756         QList *redist_region_count;
757         uint32_t redist0_capacity = virt_redist_capacity(vms, VIRT_GIC_REDIST);
758         uint32_t redist0_count = MIN(smp_cpus, redist0_capacity);
759 
760         nb_redist_regions = virt_gicv3_redist_region_count(vms);
761 
762         redist_region_count = qlist_new();
763         qlist_append_int(redist_region_count, redist0_count);
764         if (nb_redist_regions == 2) {
765             uint32_t redist1_capacity =
766                 virt_redist_capacity(vms, VIRT_HIGH_GIC_REDIST2);
767 
768             qlist_append_int(redist_region_count,
769                 MIN(smp_cpus - redist0_count, redist1_capacity));
770         }
771         qdev_prop_set_array(vms->gic, "redist-region-count",
772                             redist_region_count);
773 
774         if (!kvm_irqchip_in_kernel()) {
775             if (vms->tcg_its) {
776                 object_property_set_link(OBJECT(vms->gic), "sysmem",
777                                          OBJECT(mem), &error_fatal);
778                 qdev_prop_set_bit(vms->gic, "has-lpi", true);
779             }
780         }
781     } else {
782         if (!kvm_irqchip_in_kernel()) {
783             qdev_prop_set_bit(vms->gic, "has-virtualization-extensions",
784                               vms->virt);
785         }
786     }
787     gicbusdev = SYS_BUS_DEVICE(vms->gic);
788     sysbus_realize_and_unref(gicbusdev, &error_fatal);
789     sysbus_mmio_map(gicbusdev, 0, vms->memmap[VIRT_GIC_DIST].base);
790     if (vms->gic_version != VIRT_GIC_VERSION_2) {
791         sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_REDIST].base);
792         if (nb_redist_regions == 2) {
793             sysbus_mmio_map(gicbusdev, 2,
794                             vms->memmap[VIRT_HIGH_GIC_REDIST2].base);
795         }
796     } else {
797         sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_CPU].base);
798         if (vms->virt) {
799             sysbus_mmio_map(gicbusdev, 2, vms->memmap[VIRT_GIC_HYP].base);
800             sysbus_mmio_map(gicbusdev, 3, vms->memmap[VIRT_GIC_VCPU].base);
801         }
802     }
803 
804     /* Wire the outputs from each CPU's generic timer and the GICv3
805      * maintenance interrupt signal to the appropriate GIC PPI inputs,
806      * and the GIC's IRQ/FIQ/VIRQ/VFIQ interrupt outputs to the CPU's inputs.
807      */
808     for (i = 0; i < smp_cpus; i++) {
809         DeviceState *cpudev = DEVICE(qemu_get_cpu(i));
810         int intidbase = NUM_IRQS + i * GIC_INTERNAL;
811         /* Mapping from the output timer irq lines from the CPU to the
812          * GIC PPI inputs we use for the virt board.
813          */
814         const int timer_irq[] = {
815             [GTIMER_PHYS] = ARCH_TIMER_NS_EL1_IRQ,
816             [GTIMER_VIRT] = ARCH_TIMER_VIRT_IRQ,
817             [GTIMER_HYP]  = ARCH_TIMER_NS_EL2_IRQ,
818             [GTIMER_SEC]  = ARCH_TIMER_S_EL1_IRQ,
819         };
820 
821         for (unsigned irq = 0; irq < ARRAY_SIZE(timer_irq); irq++) {
822             qdev_connect_gpio_out(cpudev, irq,
823                                   qdev_get_gpio_in(vms->gic,
824                                                    intidbase + timer_irq[irq]));
825         }
826 
827         if (vms->gic_version != VIRT_GIC_VERSION_2) {
828             qemu_irq irq = qdev_get_gpio_in(vms->gic,
829                                             intidbase + ARCH_GIC_MAINT_IRQ);
830             qdev_connect_gpio_out_named(cpudev, "gicv3-maintenance-interrupt",
831                                         0, irq);
832         } else if (vms->virt) {
833             qemu_irq irq = qdev_get_gpio_in(vms->gic,
834                                             intidbase + ARCH_GIC_MAINT_IRQ);
835             sysbus_connect_irq(gicbusdev, i + 4 * smp_cpus, irq);
836         }
837 
838         qdev_connect_gpio_out_named(cpudev, "pmu-interrupt", 0,
839                                     qdev_get_gpio_in(vms->gic, intidbase
840                                                      + VIRTUAL_PMU_IRQ));
841 
842         sysbus_connect_irq(gicbusdev, i, qdev_get_gpio_in(cpudev, ARM_CPU_IRQ));
843         sysbus_connect_irq(gicbusdev, i + smp_cpus,
844                            qdev_get_gpio_in(cpudev, ARM_CPU_FIQ));
845         sysbus_connect_irq(gicbusdev, i + 2 * smp_cpus,
846                            qdev_get_gpio_in(cpudev, ARM_CPU_VIRQ));
847         sysbus_connect_irq(gicbusdev, i + 3 * smp_cpus,
848                            qdev_get_gpio_in(cpudev, ARM_CPU_VFIQ));
849     }
850 
851     fdt_add_gic_node(vms);
852 
853     if (vms->gic_version != VIRT_GIC_VERSION_2 && vms->its) {
854         create_its(vms);
855     } else if (vms->gic_version == VIRT_GIC_VERSION_2) {
856         create_v2m(vms);
857     }
858 }
859 
860 static void create_uart(const VirtMachineState *vms, int uart,
861                         MemoryRegion *mem, Chardev *chr)
862 {
863     char *nodename;
864     hwaddr base = vms->memmap[uart].base;
865     hwaddr size = vms->memmap[uart].size;
866     int irq = vms->irqmap[uart];
867     const char compat[] = "arm,pl011\0arm,primecell";
868     const char clocknames[] = "uartclk\0apb_pclk";
869     DeviceState *dev = qdev_new(TYPE_PL011);
870     SysBusDevice *s = SYS_BUS_DEVICE(dev);
871     MachineState *ms = MACHINE(vms);
872 
873     qdev_prop_set_chr(dev, "chardev", chr);
874     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
875     memory_region_add_subregion(mem, base,
876                                 sysbus_mmio_get_region(s, 0));
877     sysbus_connect_irq(s, 0, qdev_get_gpio_in(vms->gic, irq));
878 
879     nodename = g_strdup_printf("/pl011@%" PRIx64, base);
880     qemu_fdt_add_subnode(ms->fdt, nodename);
881     /* Note that we can't use setprop_string because of the embedded NUL */
882     qemu_fdt_setprop(ms->fdt, nodename, "compatible",
883                          compat, sizeof(compat));
884     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
885                                      2, base, 2, size);
886     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
887                                GIC_FDT_IRQ_TYPE_SPI, irq,
888                                GIC_FDT_IRQ_FLAGS_LEVEL_HI);
889     qemu_fdt_setprop_cells(ms->fdt, nodename, "clocks",
890                                vms->clock_phandle, vms->clock_phandle);
891     qemu_fdt_setprop(ms->fdt, nodename, "clock-names",
892                          clocknames, sizeof(clocknames));
893 
894     if (uart == VIRT_UART) {
895         qemu_fdt_setprop_string(ms->fdt, "/chosen", "stdout-path", nodename);
896     } else {
897         /* Mark as not usable by the normal world */
898         qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
899         qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
900 
901         qemu_fdt_setprop_string(ms->fdt, "/secure-chosen", "stdout-path",
902                                 nodename);
903     }
904 
905     g_free(nodename);
906 }
907 
908 static void create_rtc(const VirtMachineState *vms)
909 {
910     char *nodename;
911     hwaddr base = vms->memmap[VIRT_RTC].base;
912     hwaddr size = vms->memmap[VIRT_RTC].size;
913     int irq = vms->irqmap[VIRT_RTC];
914     const char compat[] = "arm,pl031\0arm,primecell";
915     MachineState *ms = MACHINE(vms);
916 
917     sysbus_create_simple("pl031", base, qdev_get_gpio_in(vms->gic, irq));
918 
919     nodename = g_strdup_printf("/pl031@%" PRIx64, base);
920     qemu_fdt_add_subnode(ms->fdt, nodename);
921     qemu_fdt_setprop(ms->fdt, nodename, "compatible", compat, sizeof(compat));
922     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
923                                  2, base, 2, size);
924     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
925                            GIC_FDT_IRQ_TYPE_SPI, irq,
926                            GIC_FDT_IRQ_FLAGS_LEVEL_HI);
927     qemu_fdt_setprop_cell(ms->fdt, nodename, "clocks", vms->clock_phandle);
928     qemu_fdt_setprop_string(ms->fdt, nodename, "clock-names", "apb_pclk");
929     g_free(nodename);
930 }
931 
932 static DeviceState *gpio_key_dev;
933 static void virt_powerdown_req(Notifier *n, void *opaque)
934 {
935     VirtMachineState *s = container_of(n, VirtMachineState, powerdown_notifier);
936 
937     if (s->acpi_dev) {
938         acpi_send_event(s->acpi_dev, ACPI_POWER_DOWN_STATUS);
939     } else {
940         /* use gpio Pin 3 for power button event */
941         qemu_set_irq(qdev_get_gpio_in(gpio_key_dev, 0), 1);
942     }
943 }
944 
945 static void create_gpio_keys(char *fdt, DeviceState *pl061_dev,
946                              uint32_t phandle)
947 {
948     gpio_key_dev = sysbus_create_simple("gpio-key", -1,
949                                         qdev_get_gpio_in(pl061_dev, 3));
950 
951     qemu_fdt_add_subnode(fdt, "/gpio-keys");
952     qemu_fdt_setprop_string(fdt, "/gpio-keys", "compatible", "gpio-keys");
953 
954     qemu_fdt_add_subnode(fdt, "/gpio-keys/poweroff");
955     qemu_fdt_setprop_string(fdt, "/gpio-keys/poweroff",
956                             "label", "GPIO Key Poweroff");
957     qemu_fdt_setprop_cell(fdt, "/gpio-keys/poweroff", "linux,code",
958                           KEY_POWER);
959     qemu_fdt_setprop_cells(fdt, "/gpio-keys/poweroff",
960                            "gpios", phandle, 3, 0);
961 }
962 
963 #define SECURE_GPIO_POWEROFF 0
964 #define SECURE_GPIO_RESET    1
965 
966 static void create_secure_gpio_pwr(char *fdt, DeviceState *pl061_dev,
967                                    uint32_t phandle)
968 {
969     DeviceState *gpio_pwr_dev;
970 
971     /* gpio-pwr */
972     gpio_pwr_dev = sysbus_create_simple("gpio-pwr", -1, NULL);
973 
974     /* connect secure pl061 to gpio-pwr */
975     qdev_connect_gpio_out(pl061_dev, SECURE_GPIO_RESET,
976                           qdev_get_gpio_in_named(gpio_pwr_dev, "reset", 0));
977     qdev_connect_gpio_out(pl061_dev, SECURE_GPIO_POWEROFF,
978                           qdev_get_gpio_in_named(gpio_pwr_dev, "shutdown", 0));
979 
980     qemu_fdt_add_subnode(fdt, "/gpio-poweroff");
981     qemu_fdt_setprop_string(fdt, "/gpio-poweroff", "compatible",
982                             "gpio-poweroff");
983     qemu_fdt_setprop_cells(fdt, "/gpio-poweroff",
984                            "gpios", phandle, SECURE_GPIO_POWEROFF, 0);
985     qemu_fdt_setprop_string(fdt, "/gpio-poweroff", "status", "disabled");
986     qemu_fdt_setprop_string(fdt, "/gpio-poweroff", "secure-status",
987                             "okay");
988 
989     qemu_fdt_add_subnode(fdt, "/gpio-restart");
990     qemu_fdt_setprop_string(fdt, "/gpio-restart", "compatible",
991                             "gpio-restart");
992     qemu_fdt_setprop_cells(fdt, "/gpio-restart",
993                            "gpios", phandle, SECURE_GPIO_RESET, 0);
994     qemu_fdt_setprop_string(fdt, "/gpio-restart", "status", "disabled");
995     qemu_fdt_setprop_string(fdt, "/gpio-restart", "secure-status",
996                             "okay");
997 }
998 
999 static void create_gpio_devices(const VirtMachineState *vms, int gpio,
1000                                 MemoryRegion *mem)
1001 {
1002     char *nodename;
1003     DeviceState *pl061_dev;
1004     hwaddr base = vms->memmap[gpio].base;
1005     hwaddr size = vms->memmap[gpio].size;
1006     int irq = vms->irqmap[gpio];
1007     const char compat[] = "arm,pl061\0arm,primecell";
1008     SysBusDevice *s;
1009     MachineState *ms = MACHINE(vms);
1010 
1011     pl061_dev = qdev_new("pl061");
1012     /* Pull lines down to 0 if not driven by the PL061 */
1013     qdev_prop_set_uint32(pl061_dev, "pullups", 0);
1014     qdev_prop_set_uint32(pl061_dev, "pulldowns", 0xff);
1015     s = SYS_BUS_DEVICE(pl061_dev);
1016     sysbus_realize_and_unref(s, &error_fatal);
1017     memory_region_add_subregion(mem, base, sysbus_mmio_get_region(s, 0));
1018     sysbus_connect_irq(s, 0, qdev_get_gpio_in(vms->gic, irq));
1019 
1020     uint32_t phandle = qemu_fdt_alloc_phandle(ms->fdt);
1021     nodename = g_strdup_printf("/pl061@%" PRIx64, base);
1022     qemu_fdt_add_subnode(ms->fdt, nodename);
1023     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1024                                  2, base, 2, size);
1025     qemu_fdt_setprop(ms->fdt, nodename, "compatible", compat, sizeof(compat));
1026     qemu_fdt_setprop_cell(ms->fdt, nodename, "#gpio-cells", 2);
1027     qemu_fdt_setprop(ms->fdt, nodename, "gpio-controller", NULL, 0);
1028     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
1029                            GIC_FDT_IRQ_TYPE_SPI, irq,
1030                            GIC_FDT_IRQ_FLAGS_LEVEL_HI);
1031     qemu_fdt_setprop_cell(ms->fdt, nodename, "clocks", vms->clock_phandle);
1032     qemu_fdt_setprop_string(ms->fdt, nodename, "clock-names", "apb_pclk");
1033     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", phandle);
1034 
1035     if (gpio != VIRT_GPIO) {
1036         /* Mark as not usable by the normal world */
1037         qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
1038         qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
1039     }
1040     g_free(nodename);
1041 
1042     /* Child gpio devices */
1043     if (gpio == VIRT_GPIO) {
1044         create_gpio_keys(ms->fdt, pl061_dev, phandle);
1045     } else {
1046         create_secure_gpio_pwr(ms->fdt, pl061_dev, phandle);
1047     }
1048 }
1049 
1050 static void create_virtio_devices(const VirtMachineState *vms)
1051 {
1052     int i;
1053     hwaddr size = vms->memmap[VIRT_MMIO].size;
1054     MachineState *ms = MACHINE(vms);
1055 
1056     /* We create the transports in forwards order. Since qbus_realize()
1057      * prepends (not appends) new child buses, the incrementing loop below will
1058      * create a list of virtio-mmio buses with decreasing base addresses.
1059      *
1060      * When a -device option is processed from the command line,
1061      * qbus_find_recursive() picks the next free virtio-mmio bus in forwards
1062      * order. The upshot is that -device options in increasing command line
1063      * order are mapped to virtio-mmio buses with decreasing base addresses.
1064      *
1065      * When this code was originally written, that arrangement ensured that the
1066      * guest Linux kernel would give the lowest "name" (/dev/vda, eth0, etc) to
1067      * the first -device on the command line. (The end-to-end order is a
1068      * function of this loop, qbus_realize(), qbus_find_recursive(), and the
1069      * guest kernel's name-to-address assignment strategy.)
1070      *
1071      * Meanwhile, the kernel's traversal seems to have been reversed; see eg.
1072      * the message, if not necessarily the code, of commit 70161ff336.
1073      * Therefore the loop now establishes the inverse of the original intent.
1074      *
1075      * Unfortunately, we can't counteract the kernel change by reversing the
1076      * loop; it would break existing command lines.
1077      *
1078      * In any case, the kernel makes no guarantee about the stability of
1079      * enumeration order of virtio devices (as demonstrated by it changing
1080      * between kernel versions). For reliable and stable identification
1081      * of disks users must use UUIDs or similar mechanisms.
1082      */
1083     for (i = 0; i < NUM_VIRTIO_TRANSPORTS; i++) {
1084         int irq = vms->irqmap[VIRT_MMIO] + i;
1085         hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
1086 
1087         sysbus_create_simple("virtio-mmio", base,
1088                              qdev_get_gpio_in(vms->gic, irq));
1089     }
1090 
1091     /* We add dtb nodes in reverse order so that they appear in the finished
1092      * device tree lowest address first.
1093      *
1094      * Note that this mapping is independent of the loop above. The previous
1095      * loop influences virtio device to virtio transport assignment, whereas
1096      * this loop controls how virtio transports are laid out in the dtb.
1097      */
1098     for (i = NUM_VIRTIO_TRANSPORTS - 1; i >= 0; i--) {
1099         char *nodename;
1100         int irq = vms->irqmap[VIRT_MMIO] + i;
1101         hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
1102 
1103         nodename = g_strdup_printf("/virtio_mmio@%" PRIx64, base);
1104         qemu_fdt_add_subnode(ms->fdt, nodename);
1105         qemu_fdt_setprop_string(ms->fdt, nodename,
1106                                 "compatible", "virtio,mmio");
1107         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1108                                      2, base, 2, size);
1109         qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
1110                                GIC_FDT_IRQ_TYPE_SPI, irq,
1111                                GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
1112         qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1113         g_free(nodename);
1114     }
1115 }
1116 
1117 #define VIRT_FLASH_SECTOR_SIZE (256 * KiB)
1118 
1119 static PFlashCFI01 *virt_flash_create1(VirtMachineState *vms,
1120                                         const char *name,
1121                                         const char *alias_prop_name)
1122 {
1123     /*
1124      * Create a single flash device.  We use the same parameters as
1125      * the flash devices on the Versatile Express board.
1126      */
1127     DeviceState *dev = qdev_new(TYPE_PFLASH_CFI01);
1128 
1129     qdev_prop_set_uint64(dev, "sector-length", VIRT_FLASH_SECTOR_SIZE);
1130     qdev_prop_set_uint8(dev, "width", 4);
1131     qdev_prop_set_uint8(dev, "device-width", 2);
1132     qdev_prop_set_bit(dev, "big-endian", false);
1133     qdev_prop_set_uint16(dev, "id0", 0x89);
1134     qdev_prop_set_uint16(dev, "id1", 0x18);
1135     qdev_prop_set_uint16(dev, "id2", 0x00);
1136     qdev_prop_set_uint16(dev, "id3", 0x00);
1137     qdev_prop_set_string(dev, "name", name);
1138     object_property_add_child(OBJECT(vms), name, OBJECT(dev));
1139     object_property_add_alias(OBJECT(vms), alias_prop_name,
1140                               OBJECT(dev), "drive");
1141     return PFLASH_CFI01(dev);
1142 }
1143 
1144 static void virt_flash_create(VirtMachineState *vms)
1145 {
1146     vms->flash[0] = virt_flash_create1(vms, "virt.flash0", "pflash0");
1147     vms->flash[1] = virt_flash_create1(vms, "virt.flash1", "pflash1");
1148 }
1149 
1150 static void virt_flash_map1(PFlashCFI01 *flash,
1151                             hwaddr base, hwaddr size,
1152                             MemoryRegion *sysmem)
1153 {
1154     DeviceState *dev = DEVICE(flash);
1155 
1156     assert(QEMU_IS_ALIGNED(size, VIRT_FLASH_SECTOR_SIZE));
1157     assert(size / VIRT_FLASH_SECTOR_SIZE <= UINT32_MAX);
1158     qdev_prop_set_uint32(dev, "num-blocks", size / VIRT_FLASH_SECTOR_SIZE);
1159     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1160 
1161     memory_region_add_subregion(sysmem, base,
1162                                 sysbus_mmio_get_region(SYS_BUS_DEVICE(dev),
1163                                                        0));
1164 }
1165 
1166 static void virt_flash_map(VirtMachineState *vms,
1167                            MemoryRegion *sysmem,
1168                            MemoryRegion *secure_sysmem)
1169 {
1170     /*
1171      * Map two flash devices to fill the VIRT_FLASH space in the memmap.
1172      * sysmem is the system memory space. secure_sysmem is the secure view
1173      * of the system, and the first flash device should be made visible only
1174      * there. The second flash device is visible to both secure and nonsecure.
1175      * If sysmem == secure_sysmem this means there is no separate Secure
1176      * address space and both flash devices are generally visible.
1177      */
1178     hwaddr flashsize = vms->memmap[VIRT_FLASH].size / 2;
1179     hwaddr flashbase = vms->memmap[VIRT_FLASH].base;
1180 
1181     virt_flash_map1(vms->flash[0], flashbase, flashsize,
1182                     secure_sysmem);
1183     virt_flash_map1(vms->flash[1], flashbase + flashsize, flashsize,
1184                     sysmem);
1185 }
1186 
1187 static void virt_flash_fdt(VirtMachineState *vms,
1188                            MemoryRegion *sysmem,
1189                            MemoryRegion *secure_sysmem)
1190 {
1191     hwaddr flashsize = vms->memmap[VIRT_FLASH].size / 2;
1192     hwaddr flashbase = vms->memmap[VIRT_FLASH].base;
1193     MachineState *ms = MACHINE(vms);
1194     char *nodename;
1195 
1196     if (sysmem == secure_sysmem) {
1197         /* Report both flash devices as a single node in the DT */
1198         nodename = g_strdup_printf("/flash@%" PRIx64, flashbase);
1199         qemu_fdt_add_subnode(ms->fdt, nodename);
1200         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible", "cfi-flash");
1201         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1202                                      2, flashbase, 2, flashsize,
1203                                      2, flashbase + flashsize, 2, flashsize);
1204         qemu_fdt_setprop_cell(ms->fdt, nodename, "bank-width", 4);
1205         g_free(nodename);
1206     } else {
1207         /*
1208          * Report the devices as separate nodes so we can mark one as
1209          * only visible to the secure world.
1210          */
1211         nodename = g_strdup_printf("/secflash@%" PRIx64, flashbase);
1212         qemu_fdt_add_subnode(ms->fdt, nodename);
1213         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible", "cfi-flash");
1214         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1215                                      2, flashbase, 2, flashsize);
1216         qemu_fdt_setprop_cell(ms->fdt, nodename, "bank-width", 4);
1217         qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
1218         qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
1219         g_free(nodename);
1220 
1221         nodename = g_strdup_printf("/flash@%" PRIx64, flashbase + flashsize);
1222         qemu_fdt_add_subnode(ms->fdt, nodename);
1223         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible", "cfi-flash");
1224         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1225                                      2, flashbase + flashsize, 2, flashsize);
1226         qemu_fdt_setprop_cell(ms->fdt, nodename, "bank-width", 4);
1227         g_free(nodename);
1228     }
1229 }
1230 
1231 static bool virt_firmware_init(VirtMachineState *vms,
1232                                MemoryRegion *sysmem,
1233                                MemoryRegion *secure_sysmem)
1234 {
1235     int i;
1236     const char *bios_name;
1237     BlockBackend *pflash_blk0;
1238 
1239     /* Map legacy -drive if=pflash to machine properties */
1240     for (i = 0; i < ARRAY_SIZE(vms->flash); i++) {
1241         pflash_cfi01_legacy_drive(vms->flash[i],
1242                                   drive_get(IF_PFLASH, 0, i));
1243     }
1244 
1245     virt_flash_map(vms, sysmem, secure_sysmem);
1246 
1247     pflash_blk0 = pflash_cfi01_get_blk(vms->flash[0]);
1248 
1249     bios_name = MACHINE(vms)->firmware;
1250     if (bios_name) {
1251         char *fname;
1252         MemoryRegion *mr;
1253         int image_size;
1254 
1255         if (pflash_blk0) {
1256             error_report("The contents of the first flash device may be "
1257                          "specified with -bios or with -drive if=pflash... "
1258                          "but you cannot use both options at once");
1259             exit(1);
1260         }
1261 
1262         /* Fall back to -bios */
1263 
1264         fname = qemu_find_file(QEMU_FILE_TYPE_BIOS, bios_name);
1265         if (!fname) {
1266             error_report("Could not find ROM image '%s'", bios_name);
1267             exit(1);
1268         }
1269         mr = sysbus_mmio_get_region(SYS_BUS_DEVICE(vms->flash[0]), 0);
1270         image_size = load_image_mr(fname, mr);
1271         g_free(fname);
1272         if (image_size < 0) {
1273             error_report("Could not load ROM image '%s'", bios_name);
1274             exit(1);
1275         }
1276     }
1277 
1278     return pflash_blk0 || bios_name;
1279 }
1280 
1281 static FWCfgState *create_fw_cfg(const VirtMachineState *vms, AddressSpace *as)
1282 {
1283     MachineState *ms = MACHINE(vms);
1284     hwaddr base = vms->memmap[VIRT_FW_CFG].base;
1285     hwaddr size = vms->memmap[VIRT_FW_CFG].size;
1286     FWCfgState *fw_cfg;
1287     char *nodename;
1288 
1289     fw_cfg = fw_cfg_init_mem_wide(base + 8, base, 8, base + 16, as);
1290     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, (uint16_t)ms->smp.cpus);
1291 
1292     nodename = g_strdup_printf("/fw-cfg@%" PRIx64, base);
1293     qemu_fdt_add_subnode(ms->fdt, nodename);
1294     qemu_fdt_setprop_string(ms->fdt, nodename,
1295                             "compatible", "qemu,fw-cfg-mmio");
1296     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1297                                  2, base, 2, size);
1298     qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1299     g_free(nodename);
1300     return fw_cfg;
1301 }
1302 
1303 static void create_pcie_irq_map(const MachineState *ms,
1304                                 uint32_t gic_phandle,
1305                                 int first_irq, const char *nodename)
1306 {
1307     int devfn, pin;
1308     uint32_t full_irq_map[4 * 4 * 10] = { 0 };
1309     uint32_t *irq_map = full_irq_map;
1310 
1311     for (devfn = 0; devfn <= 0x18; devfn += 0x8) {
1312         for (pin = 0; pin < 4; pin++) {
1313             int irq_type = GIC_FDT_IRQ_TYPE_SPI;
1314             int irq_nr = first_irq + ((pin + PCI_SLOT(devfn)) % PCI_NUM_PINS);
1315             int irq_level = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
1316             int i;
1317 
1318             uint32_t map[] = {
1319                 devfn << 8, 0, 0,                           /* devfn */
1320                 pin + 1,                                    /* PCI pin */
1321                 gic_phandle, 0, 0, irq_type, irq_nr, irq_level }; /* GIC irq */
1322 
1323             /* Convert map to big endian */
1324             for (i = 0; i < 10; i++) {
1325                 irq_map[i] = cpu_to_be32(map[i]);
1326             }
1327             irq_map += 10;
1328         }
1329     }
1330 
1331     qemu_fdt_setprop(ms->fdt, nodename, "interrupt-map",
1332                      full_irq_map, sizeof(full_irq_map));
1333 
1334     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupt-map-mask",
1335                            cpu_to_be16(PCI_DEVFN(3, 0)), /* Slot 3 */
1336                            0, 0,
1337                            0x7           /* PCI irq */);
1338 }
1339 
1340 static void create_smmu(const VirtMachineState *vms,
1341                         PCIBus *bus)
1342 {
1343     char *node;
1344     const char compat[] = "arm,smmu-v3";
1345     int irq =  vms->irqmap[VIRT_SMMU];
1346     int i;
1347     hwaddr base = vms->memmap[VIRT_SMMU].base;
1348     hwaddr size = vms->memmap[VIRT_SMMU].size;
1349     const char irq_names[] = "eventq\0priq\0cmdq-sync\0gerror";
1350     DeviceState *dev;
1351     MachineState *ms = MACHINE(vms);
1352 
1353     if (vms->iommu != VIRT_IOMMU_SMMUV3 || !vms->iommu_phandle) {
1354         return;
1355     }
1356 
1357     dev = qdev_new(TYPE_ARM_SMMUV3);
1358 
1359     object_property_set_link(OBJECT(dev), "primary-bus", OBJECT(bus),
1360                              &error_abort);
1361     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1362     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, base);
1363     for (i = 0; i < NUM_SMMU_IRQS; i++) {
1364         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
1365                            qdev_get_gpio_in(vms->gic, irq + i));
1366     }
1367 
1368     node = g_strdup_printf("/smmuv3@%" PRIx64, base);
1369     qemu_fdt_add_subnode(ms->fdt, node);
1370     qemu_fdt_setprop(ms->fdt, node, "compatible", compat, sizeof(compat));
1371     qemu_fdt_setprop_sized_cells(ms->fdt, node, "reg", 2, base, 2, size);
1372 
1373     qemu_fdt_setprop_cells(ms->fdt, node, "interrupts",
1374             GIC_FDT_IRQ_TYPE_SPI, irq    , GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1375             GIC_FDT_IRQ_TYPE_SPI, irq + 1, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1376             GIC_FDT_IRQ_TYPE_SPI, irq + 2, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1377             GIC_FDT_IRQ_TYPE_SPI, irq + 3, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
1378 
1379     qemu_fdt_setprop(ms->fdt, node, "interrupt-names", irq_names,
1380                      sizeof(irq_names));
1381 
1382     qemu_fdt_setprop(ms->fdt, node, "dma-coherent", NULL, 0);
1383 
1384     qemu_fdt_setprop_cell(ms->fdt, node, "#iommu-cells", 1);
1385 
1386     qemu_fdt_setprop_cell(ms->fdt, node, "phandle", vms->iommu_phandle);
1387     g_free(node);
1388 }
1389 
1390 static void create_virtio_iommu_dt_bindings(VirtMachineState *vms)
1391 {
1392     const char compat[] = "virtio,pci-iommu\0pci1af4,1057";
1393     uint16_t bdf = vms->virtio_iommu_bdf;
1394     MachineState *ms = MACHINE(vms);
1395     char *node;
1396 
1397     vms->iommu_phandle = qemu_fdt_alloc_phandle(ms->fdt);
1398 
1399     node = g_strdup_printf("%s/virtio_iommu@%x,%x", vms->pciehb_nodename,
1400                            PCI_SLOT(bdf), PCI_FUNC(bdf));
1401     qemu_fdt_add_subnode(ms->fdt, node);
1402     qemu_fdt_setprop(ms->fdt, node, "compatible", compat, sizeof(compat));
1403     qemu_fdt_setprop_sized_cells(ms->fdt, node, "reg",
1404                                  1, bdf << 8, 1, 0, 1, 0,
1405                                  1, 0, 1, 0);
1406 
1407     qemu_fdt_setprop_cell(ms->fdt, node, "#iommu-cells", 1);
1408     qemu_fdt_setprop_cell(ms->fdt, node, "phandle", vms->iommu_phandle);
1409     g_free(node);
1410 
1411     qemu_fdt_setprop_cells(ms->fdt, vms->pciehb_nodename, "iommu-map",
1412                            0x0, vms->iommu_phandle, 0x0, bdf,
1413                            bdf + 1, vms->iommu_phandle, bdf + 1, 0xffff - bdf);
1414 }
1415 
1416 static void create_pcie(VirtMachineState *vms)
1417 {
1418     hwaddr base_mmio = vms->memmap[VIRT_PCIE_MMIO].base;
1419     hwaddr size_mmio = vms->memmap[VIRT_PCIE_MMIO].size;
1420     hwaddr base_mmio_high = vms->memmap[VIRT_HIGH_PCIE_MMIO].base;
1421     hwaddr size_mmio_high = vms->memmap[VIRT_HIGH_PCIE_MMIO].size;
1422     hwaddr base_pio = vms->memmap[VIRT_PCIE_PIO].base;
1423     hwaddr size_pio = vms->memmap[VIRT_PCIE_PIO].size;
1424     hwaddr base_ecam, size_ecam;
1425     hwaddr base = base_mmio;
1426     int nr_pcie_buses;
1427     int irq = vms->irqmap[VIRT_PCIE];
1428     MemoryRegion *mmio_alias;
1429     MemoryRegion *mmio_reg;
1430     MemoryRegion *ecam_alias;
1431     MemoryRegion *ecam_reg;
1432     DeviceState *dev;
1433     char *nodename;
1434     int i, ecam_id;
1435     PCIHostState *pci;
1436     MachineState *ms = MACHINE(vms);
1437     MachineClass *mc = MACHINE_GET_CLASS(ms);
1438 
1439     dev = qdev_new(TYPE_GPEX_HOST);
1440     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1441 
1442     ecam_id = VIRT_ECAM_ID(vms->highmem_ecam);
1443     base_ecam = vms->memmap[ecam_id].base;
1444     size_ecam = vms->memmap[ecam_id].size;
1445     nr_pcie_buses = size_ecam / PCIE_MMCFG_SIZE_MIN;
1446     /* Map only the first size_ecam bytes of ECAM space */
1447     ecam_alias = g_new0(MemoryRegion, 1);
1448     ecam_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 0);
1449     memory_region_init_alias(ecam_alias, OBJECT(dev), "pcie-ecam",
1450                              ecam_reg, 0, size_ecam);
1451     memory_region_add_subregion(get_system_memory(), base_ecam, ecam_alias);
1452 
1453     /* Map the MMIO window into system address space so as to expose
1454      * the section of PCI MMIO space which starts at the same base address
1455      * (ie 1:1 mapping for that part of PCI MMIO space visible through
1456      * the window).
1457      */
1458     mmio_alias = g_new0(MemoryRegion, 1);
1459     mmio_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 1);
1460     memory_region_init_alias(mmio_alias, OBJECT(dev), "pcie-mmio",
1461                              mmio_reg, base_mmio, size_mmio);
1462     memory_region_add_subregion(get_system_memory(), base_mmio, mmio_alias);
1463 
1464     if (vms->highmem_mmio) {
1465         /* Map high MMIO space */
1466         MemoryRegion *high_mmio_alias = g_new0(MemoryRegion, 1);
1467 
1468         memory_region_init_alias(high_mmio_alias, OBJECT(dev), "pcie-mmio-high",
1469                                  mmio_reg, base_mmio_high, size_mmio_high);
1470         memory_region_add_subregion(get_system_memory(), base_mmio_high,
1471                                     high_mmio_alias);
1472     }
1473 
1474     /* Map IO port space */
1475     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 2, base_pio);
1476 
1477     for (i = 0; i < GPEX_NUM_IRQS; i++) {
1478         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
1479                            qdev_get_gpio_in(vms->gic, irq + i));
1480         gpex_set_irq_num(GPEX_HOST(dev), i, irq + i);
1481     }
1482 
1483     pci = PCI_HOST_BRIDGE(dev);
1484     pci->bypass_iommu = vms->default_bus_bypass_iommu;
1485     vms->bus = pci->bus;
1486     if (vms->bus) {
1487         for (i = 0; i < nb_nics; i++) {
1488             pci_nic_init_nofail(&nd_table[i], pci->bus, mc->default_nic, NULL);
1489         }
1490     }
1491 
1492     nodename = vms->pciehb_nodename = g_strdup_printf("/pcie@%" PRIx64, base);
1493     qemu_fdt_add_subnode(ms->fdt, nodename);
1494     qemu_fdt_setprop_string(ms->fdt, nodename,
1495                             "compatible", "pci-host-ecam-generic");
1496     qemu_fdt_setprop_string(ms->fdt, nodename, "device_type", "pci");
1497     qemu_fdt_setprop_cell(ms->fdt, nodename, "#address-cells", 3);
1498     qemu_fdt_setprop_cell(ms->fdt, nodename, "#size-cells", 2);
1499     qemu_fdt_setprop_cell(ms->fdt, nodename, "linux,pci-domain", 0);
1500     qemu_fdt_setprop_cells(ms->fdt, nodename, "bus-range", 0,
1501                            nr_pcie_buses - 1);
1502     qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1503 
1504     if (vms->msi_phandle) {
1505         qemu_fdt_setprop_cells(ms->fdt, nodename, "msi-map",
1506                                0, vms->msi_phandle, 0, 0x10000);
1507     }
1508 
1509     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1510                                  2, base_ecam, 2, size_ecam);
1511 
1512     if (vms->highmem_mmio) {
1513         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "ranges",
1514                                      1, FDT_PCI_RANGE_IOPORT, 2, 0,
1515                                      2, base_pio, 2, size_pio,
1516                                      1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1517                                      2, base_mmio, 2, size_mmio,
1518                                      1, FDT_PCI_RANGE_MMIO_64BIT,
1519                                      2, base_mmio_high,
1520                                      2, base_mmio_high, 2, size_mmio_high);
1521     } else {
1522         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "ranges",
1523                                      1, FDT_PCI_RANGE_IOPORT, 2, 0,
1524                                      2, base_pio, 2, size_pio,
1525                                      1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1526                                      2, base_mmio, 2, size_mmio);
1527     }
1528 
1529     qemu_fdt_setprop_cell(ms->fdt, nodename, "#interrupt-cells", 1);
1530     create_pcie_irq_map(ms, vms->gic_phandle, irq, nodename);
1531 
1532     if (vms->iommu) {
1533         vms->iommu_phandle = qemu_fdt_alloc_phandle(ms->fdt);
1534 
1535         switch (vms->iommu) {
1536         case VIRT_IOMMU_SMMUV3:
1537             create_smmu(vms, vms->bus);
1538             qemu_fdt_setprop_cells(ms->fdt, nodename, "iommu-map",
1539                                    0x0, vms->iommu_phandle, 0x0, 0x10000);
1540             break;
1541         default:
1542             g_assert_not_reached();
1543         }
1544     }
1545 }
1546 
1547 static void create_platform_bus(VirtMachineState *vms)
1548 {
1549     DeviceState *dev;
1550     SysBusDevice *s;
1551     int i;
1552     MemoryRegion *sysmem = get_system_memory();
1553 
1554     dev = qdev_new(TYPE_PLATFORM_BUS_DEVICE);
1555     dev->id = g_strdup(TYPE_PLATFORM_BUS_DEVICE);
1556     qdev_prop_set_uint32(dev, "num_irqs", PLATFORM_BUS_NUM_IRQS);
1557     qdev_prop_set_uint32(dev, "mmio_size", vms->memmap[VIRT_PLATFORM_BUS].size);
1558     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1559     vms->platform_bus_dev = dev;
1560 
1561     s = SYS_BUS_DEVICE(dev);
1562     for (i = 0; i < PLATFORM_BUS_NUM_IRQS; i++) {
1563         int irq = vms->irqmap[VIRT_PLATFORM_BUS] + i;
1564         sysbus_connect_irq(s, i, qdev_get_gpio_in(vms->gic, irq));
1565     }
1566 
1567     memory_region_add_subregion(sysmem,
1568                                 vms->memmap[VIRT_PLATFORM_BUS].base,
1569                                 sysbus_mmio_get_region(s, 0));
1570 }
1571 
1572 static void create_tag_ram(MemoryRegion *tag_sysmem,
1573                            hwaddr base, hwaddr size,
1574                            const char *name)
1575 {
1576     MemoryRegion *tagram = g_new(MemoryRegion, 1);
1577 
1578     memory_region_init_ram(tagram, NULL, name, size / 32, &error_fatal);
1579     memory_region_add_subregion(tag_sysmem, base / 32, tagram);
1580 }
1581 
1582 static void create_secure_ram(VirtMachineState *vms,
1583                               MemoryRegion *secure_sysmem,
1584                               MemoryRegion *secure_tag_sysmem)
1585 {
1586     MemoryRegion *secram = g_new(MemoryRegion, 1);
1587     char *nodename;
1588     hwaddr base = vms->memmap[VIRT_SECURE_MEM].base;
1589     hwaddr size = vms->memmap[VIRT_SECURE_MEM].size;
1590     MachineState *ms = MACHINE(vms);
1591 
1592     memory_region_init_ram(secram, NULL, "virt.secure-ram", size,
1593                            &error_fatal);
1594     memory_region_add_subregion(secure_sysmem, base, secram);
1595 
1596     nodename = g_strdup_printf("/secram@%" PRIx64, base);
1597     qemu_fdt_add_subnode(ms->fdt, nodename);
1598     qemu_fdt_setprop_string(ms->fdt, nodename, "device_type", "memory");
1599     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg", 2, base, 2, size);
1600     qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
1601     qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
1602 
1603     if (secure_tag_sysmem) {
1604         create_tag_ram(secure_tag_sysmem, base, size, "mach-virt.secure-tag");
1605     }
1606 
1607     g_free(nodename);
1608 }
1609 
1610 static void *machvirt_dtb(const struct arm_boot_info *binfo, int *fdt_size)
1611 {
1612     const VirtMachineState *board = container_of(binfo, VirtMachineState,
1613                                                  bootinfo);
1614     MachineState *ms = MACHINE(board);
1615 
1616 
1617     *fdt_size = board->fdt_size;
1618     return ms->fdt;
1619 }
1620 
1621 static void virt_build_smbios(VirtMachineState *vms)
1622 {
1623     MachineClass *mc = MACHINE_GET_CLASS(vms);
1624     MachineState *ms = MACHINE(vms);
1625     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1626     uint8_t *smbios_tables, *smbios_anchor;
1627     size_t smbios_tables_len, smbios_anchor_len;
1628     struct smbios_phys_mem_area mem_array;
1629     const char *product = "QEMU Virtual Machine";
1630 
1631     if (kvm_enabled()) {
1632         product = "KVM Virtual Machine";
1633     }
1634 
1635     smbios_set_defaults("QEMU", product,
1636                         vmc->smbios_old_sys_ver ? "1.0" : mc->name, false,
1637                         true, SMBIOS_ENTRY_POINT_TYPE_64);
1638 
1639     /* build the array of physical mem area from base_memmap */
1640     mem_array.address = vms->memmap[VIRT_MEM].base;
1641     mem_array.length = ms->ram_size;
1642 
1643     smbios_get_tables(ms, &mem_array, 1,
1644                       &smbios_tables, &smbios_tables_len,
1645                       &smbios_anchor, &smbios_anchor_len,
1646                       &error_fatal);
1647 
1648     if (smbios_anchor) {
1649         fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-tables",
1650                         smbios_tables, smbios_tables_len);
1651         fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-anchor",
1652                         smbios_anchor, smbios_anchor_len);
1653     }
1654 }
1655 
1656 static
1657 void virt_machine_done(Notifier *notifier, void *data)
1658 {
1659     VirtMachineState *vms = container_of(notifier, VirtMachineState,
1660                                          machine_done);
1661     MachineState *ms = MACHINE(vms);
1662     ARMCPU *cpu = ARM_CPU(first_cpu);
1663     struct arm_boot_info *info = &vms->bootinfo;
1664     AddressSpace *as = arm_boot_address_space(cpu, info);
1665 
1666     /*
1667      * If the user provided a dtb, we assume the dynamic sysbus nodes
1668      * already are integrated there. This corresponds to a use case where
1669      * the dynamic sysbus nodes are complex and their generation is not yet
1670      * supported. In that case the user can take charge of the guest dt
1671      * while qemu takes charge of the qom stuff.
1672      */
1673     if (info->dtb_filename == NULL) {
1674         platform_bus_add_all_fdt_nodes(ms->fdt, "/intc",
1675                                        vms->memmap[VIRT_PLATFORM_BUS].base,
1676                                        vms->memmap[VIRT_PLATFORM_BUS].size,
1677                                        vms->irqmap[VIRT_PLATFORM_BUS]);
1678     }
1679     if (arm_load_dtb(info->dtb_start, info, info->dtb_limit, as, ms) < 0) {
1680         exit(1);
1681     }
1682 
1683     fw_cfg_add_extra_pci_roots(vms->bus, vms->fw_cfg);
1684 
1685     virt_acpi_setup(vms);
1686     virt_build_smbios(vms);
1687 }
1688 
1689 static uint64_t virt_cpu_mp_affinity(VirtMachineState *vms, int idx)
1690 {
1691     uint8_t clustersz = ARM_DEFAULT_CPUS_PER_CLUSTER;
1692     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1693 
1694     if (!vmc->disallow_affinity_adjustment) {
1695         /* Adjust MPIDR like 64-bit KVM hosts, which incorporate the
1696          * GIC's target-list limitations. 32-bit KVM hosts currently
1697          * always create clusters of 4 CPUs, but that is expected to
1698          * change when they gain support for gicv3. When KVM is enabled
1699          * it will override the changes we make here, therefore our
1700          * purposes are to make TCG consistent (with 64-bit KVM hosts)
1701          * and to improve SGI efficiency.
1702          */
1703         if (vms->gic_version == VIRT_GIC_VERSION_2) {
1704             clustersz = GIC_TARGETLIST_BITS;
1705         } else {
1706             clustersz = GICV3_TARGETLIST_BITS;
1707         }
1708     }
1709     return arm_cpu_mp_affinity(idx, clustersz);
1710 }
1711 
1712 static inline bool *virt_get_high_memmap_enabled(VirtMachineState *vms,
1713                                                  int index)
1714 {
1715     bool *enabled_array[] = {
1716         &vms->highmem_redists,
1717         &vms->highmem_ecam,
1718         &vms->highmem_mmio,
1719     };
1720 
1721     assert(ARRAY_SIZE(extended_memmap) - VIRT_LOWMEMMAP_LAST ==
1722            ARRAY_SIZE(enabled_array));
1723     assert(index - VIRT_LOWMEMMAP_LAST < ARRAY_SIZE(enabled_array));
1724 
1725     return enabled_array[index - VIRT_LOWMEMMAP_LAST];
1726 }
1727 
1728 static void virt_set_high_memmap(VirtMachineState *vms,
1729                                  hwaddr base, int pa_bits)
1730 {
1731     hwaddr region_base, region_size;
1732     bool *region_enabled, fits;
1733     int i;
1734 
1735     for (i = VIRT_LOWMEMMAP_LAST; i < ARRAY_SIZE(extended_memmap); i++) {
1736         region_enabled = virt_get_high_memmap_enabled(vms, i);
1737         region_base = ROUND_UP(base, extended_memmap[i].size);
1738         region_size = extended_memmap[i].size;
1739 
1740         vms->memmap[i].base = region_base;
1741         vms->memmap[i].size = region_size;
1742 
1743         /*
1744          * Check each device to see if it fits in the PA space,
1745          * moving highest_gpa as we go. For compatibility, move
1746          * highest_gpa for disabled fitting devices as well, if
1747          * the compact layout has been disabled.
1748          *
1749          * For each device that doesn't fit, disable it.
1750          */
1751         fits = (region_base + region_size) <= BIT_ULL(pa_bits);
1752         *region_enabled &= fits;
1753         if (vms->highmem_compact && !*region_enabled) {
1754             continue;
1755         }
1756 
1757         base = region_base + region_size;
1758         if (fits) {
1759             vms->highest_gpa = base - 1;
1760         }
1761     }
1762 }
1763 
1764 static void virt_set_memmap(VirtMachineState *vms, int pa_bits)
1765 {
1766     MachineState *ms = MACHINE(vms);
1767     hwaddr base, device_memory_base, device_memory_size, memtop;
1768     int i;
1769 
1770     vms->memmap = extended_memmap;
1771 
1772     for (i = 0; i < ARRAY_SIZE(base_memmap); i++) {
1773         vms->memmap[i] = base_memmap[i];
1774     }
1775 
1776     if (ms->ram_slots > ACPI_MAX_RAM_SLOTS) {
1777         error_report("unsupported number of memory slots: %"PRIu64,
1778                      ms->ram_slots);
1779         exit(EXIT_FAILURE);
1780     }
1781 
1782     /*
1783      * !highmem is exactly the same as limiting the PA space to 32bit,
1784      * irrespective of the underlying capabilities of the HW.
1785      */
1786     if (!vms->highmem) {
1787         pa_bits = 32;
1788     }
1789 
1790     /*
1791      * We compute the base of the high IO region depending on the
1792      * amount of initial and device memory. The device memory start/size
1793      * is aligned on 1GiB. We never put the high IO region below 256GiB
1794      * so that if maxram_size is < 255GiB we keep the legacy memory map.
1795      * The device region size assumes 1GiB page max alignment per slot.
1796      */
1797     device_memory_base =
1798         ROUND_UP(vms->memmap[VIRT_MEM].base + ms->ram_size, GiB);
1799     device_memory_size = ms->maxram_size - ms->ram_size + ms->ram_slots * GiB;
1800 
1801     /* Base address of the high IO region */
1802     memtop = base = device_memory_base + ROUND_UP(device_memory_size, GiB);
1803     if (memtop > BIT_ULL(pa_bits)) {
1804 	    error_report("Addressing limited to %d bits, but memory exceeds it by %llu bytes\n",
1805 			 pa_bits, memtop - BIT_ULL(pa_bits));
1806         exit(EXIT_FAILURE);
1807     }
1808     if (base < device_memory_base) {
1809         error_report("maxmem/slots too huge");
1810         exit(EXIT_FAILURE);
1811     }
1812     if (base < vms->memmap[VIRT_MEM].base + LEGACY_RAMLIMIT_BYTES) {
1813         base = vms->memmap[VIRT_MEM].base + LEGACY_RAMLIMIT_BYTES;
1814     }
1815 
1816     /* We know for sure that at least the memory fits in the PA space */
1817     vms->highest_gpa = memtop - 1;
1818 
1819     virt_set_high_memmap(vms, base, pa_bits);
1820 
1821     if (device_memory_size > 0) {
1822         machine_memory_devices_init(ms, device_memory_base, device_memory_size);
1823     }
1824 }
1825 
1826 static VirtGICType finalize_gic_version_do(const char *accel_name,
1827                                            VirtGICType gic_version,
1828                                            int gics_supported,
1829                                            unsigned int max_cpus)
1830 {
1831     /* Convert host/max/nosel to GIC version number */
1832     switch (gic_version) {
1833     case VIRT_GIC_VERSION_HOST:
1834         if (!kvm_enabled()) {
1835             error_report("gic-version=host requires KVM");
1836             exit(1);
1837         }
1838 
1839         /* For KVM, gic-version=host means gic-version=max */
1840         return finalize_gic_version_do(accel_name, VIRT_GIC_VERSION_MAX,
1841                                        gics_supported, max_cpus);
1842     case VIRT_GIC_VERSION_MAX:
1843         if (gics_supported & VIRT_GIC_VERSION_4_MASK) {
1844             gic_version = VIRT_GIC_VERSION_4;
1845         } else if (gics_supported & VIRT_GIC_VERSION_3_MASK) {
1846             gic_version = VIRT_GIC_VERSION_3;
1847         } else {
1848             gic_version = VIRT_GIC_VERSION_2;
1849         }
1850         break;
1851     case VIRT_GIC_VERSION_NOSEL:
1852         if ((gics_supported & VIRT_GIC_VERSION_2_MASK) &&
1853             max_cpus <= GIC_NCPU) {
1854             gic_version = VIRT_GIC_VERSION_2;
1855         } else if (gics_supported & VIRT_GIC_VERSION_3_MASK) {
1856             /*
1857              * in case the host does not support v2 emulation or
1858              * the end-user requested more than 8 VCPUs we now default
1859              * to v3. In any case defaulting to v2 would be broken.
1860              */
1861             gic_version = VIRT_GIC_VERSION_3;
1862         } else if (max_cpus > GIC_NCPU) {
1863             error_report("%s only supports GICv2 emulation but more than 8 "
1864                          "vcpus are requested", accel_name);
1865             exit(1);
1866         }
1867         break;
1868     case VIRT_GIC_VERSION_2:
1869     case VIRT_GIC_VERSION_3:
1870     case VIRT_GIC_VERSION_4:
1871         break;
1872     }
1873 
1874     /* Check chosen version is effectively supported */
1875     switch (gic_version) {
1876     case VIRT_GIC_VERSION_2:
1877         if (!(gics_supported & VIRT_GIC_VERSION_2_MASK)) {
1878             error_report("%s does not support GICv2 emulation", accel_name);
1879             exit(1);
1880         }
1881         break;
1882     case VIRT_GIC_VERSION_3:
1883         if (!(gics_supported & VIRT_GIC_VERSION_3_MASK)) {
1884             error_report("%s does not support GICv3 emulation", accel_name);
1885             exit(1);
1886         }
1887         break;
1888     case VIRT_GIC_VERSION_4:
1889         if (!(gics_supported & VIRT_GIC_VERSION_4_MASK)) {
1890             error_report("%s does not support GICv4 emulation, is virtualization=on?",
1891                          accel_name);
1892             exit(1);
1893         }
1894         break;
1895     default:
1896         error_report("logic error in finalize_gic_version");
1897         exit(1);
1898         break;
1899     }
1900 
1901     return gic_version;
1902 }
1903 
1904 /*
1905  * finalize_gic_version - Determines the final gic_version
1906  * according to the gic-version property
1907  *
1908  * Default GIC type is v2
1909  */
1910 static void finalize_gic_version(VirtMachineState *vms)
1911 {
1912     const char *accel_name = current_accel_name();
1913     unsigned int max_cpus = MACHINE(vms)->smp.max_cpus;
1914     int gics_supported = 0;
1915 
1916     /* Determine which GIC versions the current environment supports */
1917     if (kvm_enabled() && kvm_irqchip_in_kernel()) {
1918         int probe_bitmap = kvm_arm_vgic_probe();
1919 
1920         if (!probe_bitmap) {
1921             error_report("Unable to determine GIC version supported by host");
1922             exit(1);
1923         }
1924 
1925         if (probe_bitmap & KVM_ARM_VGIC_V2) {
1926             gics_supported |= VIRT_GIC_VERSION_2_MASK;
1927         }
1928         if (probe_bitmap & KVM_ARM_VGIC_V3) {
1929             gics_supported |= VIRT_GIC_VERSION_3_MASK;
1930         }
1931     } else if (kvm_enabled() && !kvm_irqchip_in_kernel()) {
1932         /* KVM w/o kernel irqchip can only deal with GICv2 */
1933         gics_supported |= VIRT_GIC_VERSION_2_MASK;
1934         accel_name = "KVM with kernel-irqchip=off";
1935     } else if (tcg_enabled() || hvf_enabled() || qtest_enabled())  {
1936         gics_supported |= VIRT_GIC_VERSION_2_MASK;
1937         if (module_object_class_by_name("arm-gicv3")) {
1938             gics_supported |= VIRT_GIC_VERSION_3_MASK;
1939             if (vms->virt) {
1940                 /* GICv4 only makes sense if CPU has EL2 */
1941                 gics_supported |= VIRT_GIC_VERSION_4_MASK;
1942             }
1943         }
1944     } else {
1945         error_report("Unsupported accelerator, can not determine GIC support");
1946         exit(1);
1947     }
1948 
1949     /*
1950      * Then convert helpers like host/max to concrete GIC versions and ensure
1951      * the desired version is supported
1952      */
1953     vms->gic_version = finalize_gic_version_do(accel_name, vms->gic_version,
1954                                                gics_supported, max_cpus);
1955 }
1956 
1957 /*
1958  * virt_cpu_post_init() must be called after the CPUs have
1959  * been realized and the GIC has been created.
1960  */
1961 static void virt_cpu_post_init(VirtMachineState *vms, MemoryRegion *sysmem)
1962 {
1963     int max_cpus = MACHINE(vms)->smp.max_cpus;
1964     bool aarch64, pmu, steal_time;
1965     CPUState *cpu;
1966 
1967     aarch64 = object_property_get_bool(OBJECT(first_cpu), "aarch64", NULL);
1968     pmu = object_property_get_bool(OBJECT(first_cpu), "pmu", NULL);
1969     steal_time = object_property_get_bool(OBJECT(first_cpu),
1970                                           "kvm-steal-time", NULL);
1971 
1972     if (kvm_enabled()) {
1973         hwaddr pvtime_reg_base = vms->memmap[VIRT_PVTIME].base;
1974         hwaddr pvtime_reg_size = vms->memmap[VIRT_PVTIME].size;
1975 
1976         if (steal_time) {
1977             MemoryRegion *pvtime = g_new(MemoryRegion, 1);
1978             hwaddr pvtime_size = max_cpus * PVTIME_SIZE_PER_CPU;
1979 
1980             /* The memory region size must be a multiple of host page size. */
1981             pvtime_size = REAL_HOST_PAGE_ALIGN(pvtime_size);
1982 
1983             if (pvtime_size > pvtime_reg_size) {
1984                 error_report("pvtime requires a %" HWADDR_PRId
1985                              " byte memory region for %d CPUs,"
1986                              " but only %" HWADDR_PRId " has been reserved",
1987                              pvtime_size, max_cpus, pvtime_reg_size);
1988                 exit(1);
1989             }
1990 
1991             memory_region_init_ram(pvtime, NULL, "pvtime", pvtime_size, NULL);
1992             memory_region_add_subregion(sysmem, pvtime_reg_base, pvtime);
1993         }
1994 
1995         CPU_FOREACH(cpu) {
1996             if (pmu) {
1997                 assert(arm_feature(&ARM_CPU(cpu)->env, ARM_FEATURE_PMU));
1998                 if (kvm_irqchip_in_kernel()) {
1999                     kvm_arm_pmu_set_irq(cpu, VIRTUAL_PMU_IRQ);
2000                 }
2001                 kvm_arm_pmu_init(cpu);
2002             }
2003             if (steal_time) {
2004                 kvm_arm_pvtime_init(cpu, pvtime_reg_base +
2005                                          cpu->cpu_index * PVTIME_SIZE_PER_CPU);
2006             }
2007         }
2008     } else {
2009         if (aarch64 && vms->highmem) {
2010             int requested_pa_size = 64 - clz64(vms->highest_gpa);
2011             int pamax = arm_pamax(ARM_CPU(first_cpu));
2012 
2013             if (pamax < requested_pa_size) {
2014                 error_report("VCPU supports less PA bits (%d) than "
2015                              "requested by the memory map (%d)",
2016                              pamax, requested_pa_size);
2017                 exit(1);
2018             }
2019         }
2020     }
2021 }
2022 
2023 static void machvirt_init(MachineState *machine)
2024 {
2025     VirtMachineState *vms = VIRT_MACHINE(machine);
2026     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(machine);
2027     MachineClass *mc = MACHINE_GET_CLASS(machine);
2028     const CPUArchIdList *possible_cpus;
2029     MemoryRegion *sysmem = get_system_memory();
2030     MemoryRegion *secure_sysmem = NULL;
2031     MemoryRegion *tag_sysmem = NULL;
2032     MemoryRegion *secure_tag_sysmem = NULL;
2033     int n, virt_max_cpus;
2034     bool firmware_loaded;
2035     bool aarch64 = true;
2036     bool has_ged = !vmc->no_ged;
2037     unsigned int smp_cpus = machine->smp.cpus;
2038     unsigned int max_cpus = machine->smp.max_cpus;
2039 
2040     if (!cpu_type_valid(machine->cpu_type)) {
2041         error_report("mach-virt: CPU type %s not supported", machine->cpu_type);
2042         exit(1);
2043     }
2044 
2045     possible_cpus = mc->possible_cpu_arch_ids(machine);
2046 
2047     /*
2048      * In accelerated mode, the memory map is computed earlier in kvm_type()
2049      * to create a VM with the right number of IPA bits.
2050      */
2051     if (!vms->memmap) {
2052         Object *cpuobj;
2053         ARMCPU *armcpu;
2054         int pa_bits;
2055 
2056         /*
2057          * Instantiate a temporary CPU object to find out about what
2058          * we are about to deal with. Once this is done, get rid of
2059          * the object.
2060          */
2061         cpuobj = object_new(possible_cpus->cpus[0].type);
2062         armcpu = ARM_CPU(cpuobj);
2063 
2064         pa_bits = arm_pamax(armcpu);
2065 
2066         object_unref(cpuobj);
2067 
2068         virt_set_memmap(vms, pa_bits);
2069     }
2070 
2071     /* We can probe only here because during property set
2072      * KVM is not available yet
2073      */
2074     finalize_gic_version(vms);
2075 
2076     if (vms->secure) {
2077         /*
2078          * The Secure view of the world is the same as the NonSecure,
2079          * but with a few extra devices. Create it as a container region
2080          * containing the system memory at low priority; any secure-only
2081          * devices go in at higher priority and take precedence.
2082          */
2083         secure_sysmem = g_new(MemoryRegion, 1);
2084         memory_region_init(secure_sysmem, OBJECT(machine), "secure-memory",
2085                            UINT64_MAX);
2086         memory_region_add_subregion_overlap(secure_sysmem, 0, sysmem, -1);
2087     }
2088 
2089     firmware_loaded = virt_firmware_init(vms, sysmem,
2090                                          secure_sysmem ?: sysmem);
2091 
2092     /* If we have an EL3 boot ROM then the assumption is that it will
2093      * implement PSCI itself, so disable QEMU's internal implementation
2094      * so it doesn't get in the way. Instead of starting secondary
2095      * CPUs in PSCI powerdown state we will start them all running and
2096      * let the boot ROM sort them out.
2097      * The usual case is that we do use QEMU's PSCI implementation;
2098      * if the guest has EL2 then we will use SMC as the conduit,
2099      * and otherwise we will use HVC (for backwards compatibility and
2100      * because if we're using KVM then we must use HVC).
2101      */
2102     if (vms->secure && firmware_loaded) {
2103         vms->psci_conduit = QEMU_PSCI_CONDUIT_DISABLED;
2104     } else if (vms->virt) {
2105         vms->psci_conduit = QEMU_PSCI_CONDUIT_SMC;
2106     } else {
2107         vms->psci_conduit = QEMU_PSCI_CONDUIT_HVC;
2108     }
2109 
2110     /*
2111      * The maximum number of CPUs depends on the GIC version, or on how
2112      * many redistributors we can fit into the memory map (which in turn
2113      * depends on whether this is a GICv3 or v4).
2114      */
2115     if (vms->gic_version == VIRT_GIC_VERSION_2) {
2116         virt_max_cpus = GIC_NCPU;
2117     } else {
2118         virt_max_cpus = virt_redist_capacity(vms, VIRT_GIC_REDIST);
2119         if (vms->highmem_redists) {
2120             virt_max_cpus += virt_redist_capacity(vms, VIRT_HIGH_GIC_REDIST2);
2121         }
2122     }
2123 
2124     if (max_cpus > virt_max_cpus) {
2125         error_report("Number of SMP CPUs requested (%d) exceeds max CPUs "
2126                      "supported by machine 'mach-virt' (%d)",
2127                      max_cpus, virt_max_cpus);
2128         if (vms->gic_version != VIRT_GIC_VERSION_2 && !vms->highmem_redists) {
2129             error_printf("Try 'highmem-redists=on' for more CPUs\n");
2130         }
2131 
2132         exit(1);
2133     }
2134 
2135     if (vms->secure && (kvm_enabled() || hvf_enabled())) {
2136         error_report("mach-virt: %s does not support providing "
2137                      "Security extensions (TrustZone) to the guest CPU",
2138                      current_accel_name());
2139         exit(1);
2140     }
2141 
2142     if (vms->virt && (kvm_enabled() || hvf_enabled())) {
2143         error_report("mach-virt: %s does not support providing "
2144                      "Virtualization extensions to the guest CPU",
2145                      current_accel_name());
2146         exit(1);
2147     }
2148 
2149     if (vms->mte && (kvm_enabled() || hvf_enabled())) {
2150         error_report("mach-virt: %s does not support providing "
2151                      "MTE to the guest CPU",
2152                      current_accel_name());
2153         exit(1);
2154     }
2155 
2156     create_fdt(vms);
2157 
2158     assert(possible_cpus->len == max_cpus);
2159     for (n = 0; n < possible_cpus->len; n++) {
2160         Object *cpuobj;
2161         CPUState *cs;
2162 
2163         if (n >= smp_cpus) {
2164             break;
2165         }
2166 
2167         cpuobj = object_new(possible_cpus->cpus[n].type);
2168         object_property_set_int(cpuobj, "mp-affinity",
2169                                 possible_cpus->cpus[n].arch_id, NULL);
2170 
2171         cs = CPU(cpuobj);
2172         cs->cpu_index = n;
2173 
2174         numa_cpu_pre_plug(&possible_cpus->cpus[cs->cpu_index], DEVICE(cpuobj),
2175                           &error_fatal);
2176 
2177         aarch64 &= object_property_get_bool(cpuobj, "aarch64", NULL);
2178 
2179         if (!vms->secure) {
2180             object_property_set_bool(cpuobj, "has_el3", false, NULL);
2181         }
2182 
2183         if (!vms->virt && object_property_find(cpuobj, "has_el2")) {
2184             object_property_set_bool(cpuobj, "has_el2", false, NULL);
2185         }
2186 
2187         if (vmc->kvm_no_adjvtime &&
2188             object_property_find(cpuobj, "kvm-no-adjvtime")) {
2189             object_property_set_bool(cpuobj, "kvm-no-adjvtime", true, NULL);
2190         }
2191 
2192         if (vmc->no_kvm_steal_time &&
2193             object_property_find(cpuobj, "kvm-steal-time")) {
2194             object_property_set_bool(cpuobj, "kvm-steal-time", false, NULL);
2195         }
2196 
2197         if (vmc->no_pmu && object_property_find(cpuobj, "pmu")) {
2198             object_property_set_bool(cpuobj, "pmu", false, NULL);
2199         }
2200 
2201         if (vmc->no_tcg_lpa2 && object_property_find(cpuobj, "lpa2")) {
2202             object_property_set_bool(cpuobj, "lpa2", false, NULL);
2203         }
2204 
2205         if (object_property_find(cpuobj, "reset-cbar")) {
2206             object_property_set_int(cpuobj, "reset-cbar",
2207                                     vms->memmap[VIRT_CPUPERIPHS].base,
2208                                     &error_abort);
2209         }
2210 
2211         object_property_set_link(cpuobj, "memory", OBJECT(sysmem),
2212                                  &error_abort);
2213         if (vms->secure) {
2214             object_property_set_link(cpuobj, "secure-memory",
2215                                      OBJECT(secure_sysmem), &error_abort);
2216         }
2217 
2218         if (vms->mte) {
2219             /* Create the memory region only once, but link to all cpus. */
2220             if (!tag_sysmem) {
2221                 /*
2222                  * The property exists only if MemTag is supported.
2223                  * If it is, we must allocate the ram to back that up.
2224                  */
2225                 if (!object_property_find(cpuobj, "tag-memory")) {
2226                     error_report("MTE requested, but not supported "
2227                                  "by the guest CPU");
2228                     exit(1);
2229                 }
2230 
2231                 tag_sysmem = g_new(MemoryRegion, 1);
2232                 memory_region_init(tag_sysmem, OBJECT(machine),
2233                                    "tag-memory", UINT64_MAX / 32);
2234 
2235                 if (vms->secure) {
2236                     secure_tag_sysmem = g_new(MemoryRegion, 1);
2237                     memory_region_init(secure_tag_sysmem, OBJECT(machine),
2238                                        "secure-tag-memory", UINT64_MAX / 32);
2239 
2240                     /* As with ram, secure-tag takes precedence over tag.  */
2241                     memory_region_add_subregion_overlap(secure_tag_sysmem, 0,
2242                                                         tag_sysmem, -1);
2243                 }
2244             }
2245 
2246             object_property_set_link(cpuobj, "tag-memory", OBJECT(tag_sysmem),
2247                                      &error_abort);
2248             if (vms->secure) {
2249                 object_property_set_link(cpuobj, "secure-tag-memory",
2250                                          OBJECT(secure_tag_sysmem),
2251                                          &error_abort);
2252             }
2253         }
2254 
2255         qdev_realize(DEVICE(cpuobj), NULL, &error_fatal);
2256         object_unref(cpuobj);
2257     }
2258     fdt_add_timer_nodes(vms);
2259     fdt_add_cpu_nodes(vms);
2260 
2261     memory_region_add_subregion(sysmem, vms->memmap[VIRT_MEM].base,
2262                                 machine->ram);
2263 
2264     virt_flash_fdt(vms, sysmem, secure_sysmem ?: sysmem);
2265 
2266     create_gic(vms, sysmem);
2267 
2268     virt_cpu_post_init(vms, sysmem);
2269 
2270     fdt_add_pmu_nodes(vms);
2271 
2272     create_uart(vms, VIRT_UART, sysmem, serial_hd(0));
2273 
2274     if (vms->secure) {
2275         create_secure_ram(vms, secure_sysmem, secure_tag_sysmem);
2276         create_uart(vms, VIRT_SECURE_UART, secure_sysmem, serial_hd(1));
2277     }
2278 
2279     if (tag_sysmem) {
2280         create_tag_ram(tag_sysmem, vms->memmap[VIRT_MEM].base,
2281                        machine->ram_size, "mach-virt.tag");
2282     }
2283 
2284     vms->highmem_ecam &= (!firmware_loaded || aarch64);
2285 
2286     create_rtc(vms);
2287 
2288     create_pcie(vms);
2289 
2290     if (has_ged && aarch64 && firmware_loaded && virt_is_acpi_enabled(vms)) {
2291         vms->acpi_dev = create_acpi_ged(vms);
2292     } else {
2293         create_gpio_devices(vms, VIRT_GPIO, sysmem);
2294     }
2295 
2296     if (vms->secure && !vmc->no_secure_gpio) {
2297         create_gpio_devices(vms, VIRT_SECURE_GPIO, secure_sysmem);
2298     }
2299 
2300      /* connect powerdown request */
2301      vms->powerdown_notifier.notify = virt_powerdown_req;
2302      qemu_register_powerdown_notifier(&vms->powerdown_notifier);
2303 
2304     /* Create mmio transports, so the user can create virtio backends
2305      * (which will be automatically plugged in to the transports). If
2306      * no backend is created the transport will just sit harmlessly idle.
2307      */
2308     create_virtio_devices(vms);
2309 
2310     vms->fw_cfg = create_fw_cfg(vms, &address_space_memory);
2311     rom_set_fw(vms->fw_cfg);
2312 
2313     create_platform_bus(vms);
2314 
2315     if (machine->nvdimms_state->is_enabled) {
2316         const struct AcpiGenericAddress arm_virt_nvdimm_acpi_dsmio = {
2317             .space_id = AML_AS_SYSTEM_MEMORY,
2318             .address = vms->memmap[VIRT_NVDIMM_ACPI].base,
2319             .bit_width = NVDIMM_ACPI_IO_LEN << 3
2320         };
2321 
2322         nvdimm_init_acpi_state(machine->nvdimms_state, sysmem,
2323                                arm_virt_nvdimm_acpi_dsmio,
2324                                vms->fw_cfg, OBJECT(vms));
2325     }
2326 
2327     vms->bootinfo.ram_size = machine->ram_size;
2328     vms->bootinfo.board_id = -1;
2329     vms->bootinfo.loader_start = vms->memmap[VIRT_MEM].base;
2330     vms->bootinfo.get_dtb = machvirt_dtb;
2331     vms->bootinfo.skip_dtb_autoload = true;
2332     vms->bootinfo.firmware_loaded = firmware_loaded;
2333     vms->bootinfo.psci_conduit = vms->psci_conduit;
2334     arm_load_kernel(ARM_CPU(first_cpu), machine, &vms->bootinfo);
2335 
2336     vms->machine_done.notify = virt_machine_done;
2337     qemu_add_machine_init_done_notifier(&vms->machine_done);
2338 }
2339 
2340 static bool virt_get_secure(Object *obj, Error **errp)
2341 {
2342     VirtMachineState *vms = VIRT_MACHINE(obj);
2343 
2344     return vms->secure;
2345 }
2346 
2347 static void virt_set_secure(Object *obj, bool value, Error **errp)
2348 {
2349     VirtMachineState *vms = VIRT_MACHINE(obj);
2350 
2351     vms->secure = value;
2352 }
2353 
2354 static bool virt_get_virt(Object *obj, Error **errp)
2355 {
2356     VirtMachineState *vms = VIRT_MACHINE(obj);
2357 
2358     return vms->virt;
2359 }
2360 
2361 static void virt_set_virt(Object *obj, bool value, Error **errp)
2362 {
2363     VirtMachineState *vms = VIRT_MACHINE(obj);
2364 
2365     vms->virt = value;
2366 }
2367 
2368 static bool virt_get_highmem(Object *obj, Error **errp)
2369 {
2370     VirtMachineState *vms = VIRT_MACHINE(obj);
2371 
2372     return vms->highmem;
2373 }
2374 
2375 static void virt_set_highmem(Object *obj, bool value, Error **errp)
2376 {
2377     VirtMachineState *vms = VIRT_MACHINE(obj);
2378 
2379     vms->highmem = value;
2380 }
2381 
2382 static bool virt_get_compact_highmem(Object *obj, Error **errp)
2383 {
2384     VirtMachineState *vms = VIRT_MACHINE(obj);
2385 
2386     return vms->highmem_compact;
2387 }
2388 
2389 static void virt_set_compact_highmem(Object *obj, bool value, Error **errp)
2390 {
2391     VirtMachineState *vms = VIRT_MACHINE(obj);
2392 
2393     vms->highmem_compact = value;
2394 }
2395 
2396 static bool virt_get_highmem_redists(Object *obj, Error **errp)
2397 {
2398     VirtMachineState *vms = VIRT_MACHINE(obj);
2399 
2400     return vms->highmem_redists;
2401 }
2402 
2403 static void virt_set_highmem_redists(Object *obj, bool value, Error **errp)
2404 {
2405     VirtMachineState *vms = VIRT_MACHINE(obj);
2406 
2407     vms->highmem_redists = value;
2408 }
2409 
2410 static bool virt_get_highmem_ecam(Object *obj, Error **errp)
2411 {
2412     VirtMachineState *vms = VIRT_MACHINE(obj);
2413 
2414     return vms->highmem_ecam;
2415 }
2416 
2417 static void virt_set_highmem_ecam(Object *obj, bool value, Error **errp)
2418 {
2419     VirtMachineState *vms = VIRT_MACHINE(obj);
2420 
2421     vms->highmem_ecam = value;
2422 }
2423 
2424 static bool virt_get_highmem_mmio(Object *obj, Error **errp)
2425 {
2426     VirtMachineState *vms = VIRT_MACHINE(obj);
2427 
2428     return vms->highmem_mmio;
2429 }
2430 
2431 static void virt_set_highmem_mmio(Object *obj, bool value, Error **errp)
2432 {
2433     VirtMachineState *vms = VIRT_MACHINE(obj);
2434 
2435     vms->highmem_mmio = value;
2436 }
2437 
2438 
2439 static bool virt_get_its(Object *obj, Error **errp)
2440 {
2441     VirtMachineState *vms = VIRT_MACHINE(obj);
2442 
2443     return vms->its;
2444 }
2445 
2446 static void virt_set_its(Object *obj, bool value, Error **errp)
2447 {
2448     VirtMachineState *vms = VIRT_MACHINE(obj);
2449 
2450     vms->its = value;
2451 }
2452 
2453 static bool virt_get_dtb_randomness(Object *obj, Error **errp)
2454 {
2455     VirtMachineState *vms = VIRT_MACHINE(obj);
2456 
2457     return vms->dtb_randomness;
2458 }
2459 
2460 static void virt_set_dtb_randomness(Object *obj, bool value, Error **errp)
2461 {
2462     VirtMachineState *vms = VIRT_MACHINE(obj);
2463 
2464     vms->dtb_randomness = value;
2465 }
2466 
2467 static char *virt_get_oem_id(Object *obj, Error **errp)
2468 {
2469     VirtMachineState *vms = VIRT_MACHINE(obj);
2470 
2471     return g_strdup(vms->oem_id);
2472 }
2473 
2474 static void virt_set_oem_id(Object *obj, const char *value, Error **errp)
2475 {
2476     VirtMachineState *vms = VIRT_MACHINE(obj);
2477     size_t len = strlen(value);
2478 
2479     if (len > 6) {
2480         error_setg(errp,
2481                    "User specified oem-id value is bigger than 6 bytes in size");
2482         return;
2483     }
2484 
2485     strncpy(vms->oem_id, value, 6);
2486 }
2487 
2488 static char *virt_get_oem_table_id(Object *obj, Error **errp)
2489 {
2490     VirtMachineState *vms = VIRT_MACHINE(obj);
2491 
2492     return g_strdup(vms->oem_table_id);
2493 }
2494 
2495 static void virt_set_oem_table_id(Object *obj, const char *value,
2496                                   Error **errp)
2497 {
2498     VirtMachineState *vms = VIRT_MACHINE(obj);
2499     size_t len = strlen(value);
2500 
2501     if (len > 8) {
2502         error_setg(errp,
2503                    "User specified oem-table-id value is bigger than 8 bytes in size");
2504         return;
2505     }
2506     strncpy(vms->oem_table_id, value, 8);
2507 }
2508 
2509 
2510 bool virt_is_acpi_enabled(VirtMachineState *vms)
2511 {
2512     if (vms->acpi == ON_OFF_AUTO_OFF) {
2513         return false;
2514     }
2515     return true;
2516 }
2517 
2518 static void virt_get_acpi(Object *obj, Visitor *v, const char *name,
2519                           void *opaque, Error **errp)
2520 {
2521     VirtMachineState *vms = VIRT_MACHINE(obj);
2522     OnOffAuto acpi = vms->acpi;
2523 
2524     visit_type_OnOffAuto(v, name, &acpi, errp);
2525 }
2526 
2527 static void virt_set_acpi(Object *obj, Visitor *v, const char *name,
2528                           void *opaque, Error **errp)
2529 {
2530     VirtMachineState *vms = VIRT_MACHINE(obj);
2531 
2532     visit_type_OnOffAuto(v, name, &vms->acpi, errp);
2533 }
2534 
2535 static bool virt_get_ras(Object *obj, Error **errp)
2536 {
2537     VirtMachineState *vms = VIRT_MACHINE(obj);
2538 
2539     return vms->ras;
2540 }
2541 
2542 static void virt_set_ras(Object *obj, bool value, Error **errp)
2543 {
2544     VirtMachineState *vms = VIRT_MACHINE(obj);
2545 
2546     vms->ras = value;
2547 }
2548 
2549 static bool virt_get_mte(Object *obj, Error **errp)
2550 {
2551     VirtMachineState *vms = VIRT_MACHINE(obj);
2552 
2553     return vms->mte;
2554 }
2555 
2556 static void virt_set_mte(Object *obj, bool value, Error **errp)
2557 {
2558     VirtMachineState *vms = VIRT_MACHINE(obj);
2559 
2560     vms->mte = value;
2561 }
2562 
2563 static char *virt_get_gic_version(Object *obj, Error **errp)
2564 {
2565     VirtMachineState *vms = VIRT_MACHINE(obj);
2566     const char *val;
2567 
2568     switch (vms->gic_version) {
2569     case VIRT_GIC_VERSION_4:
2570         val = "4";
2571         break;
2572     case VIRT_GIC_VERSION_3:
2573         val = "3";
2574         break;
2575     default:
2576         val = "2";
2577         break;
2578     }
2579     return g_strdup(val);
2580 }
2581 
2582 static void virt_set_gic_version(Object *obj, const char *value, Error **errp)
2583 {
2584     VirtMachineState *vms = VIRT_MACHINE(obj);
2585 
2586     if (!strcmp(value, "4")) {
2587         vms->gic_version = VIRT_GIC_VERSION_4;
2588     } else if (!strcmp(value, "3")) {
2589         vms->gic_version = VIRT_GIC_VERSION_3;
2590     } else if (!strcmp(value, "2")) {
2591         vms->gic_version = VIRT_GIC_VERSION_2;
2592     } else if (!strcmp(value, "host")) {
2593         vms->gic_version = VIRT_GIC_VERSION_HOST; /* Will probe later */
2594     } else if (!strcmp(value, "max")) {
2595         vms->gic_version = VIRT_GIC_VERSION_MAX; /* Will probe later */
2596     } else {
2597         error_setg(errp, "Invalid gic-version value");
2598         error_append_hint(errp, "Valid values are 3, 2, host, max.\n");
2599     }
2600 }
2601 
2602 static char *virt_get_iommu(Object *obj, Error **errp)
2603 {
2604     VirtMachineState *vms = VIRT_MACHINE(obj);
2605 
2606     switch (vms->iommu) {
2607     case VIRT_IOMMU_NONE:
2608         return g_strdup("none");
2609     case VIRT_IOMMU_SMMUV3:
2610         return g_strdup("smmuv3");
2611     default:
2612         g_assert_not_reached();
2613     }
2614 }
2615 
2616 static void virt_set_iommu(Object *obj, const char *value, Error **errp)
2617 {
2618     VirtMachineState *vms = VIRT_MACHINE(obj);
2619 
2620     if (!strcmp(value, "smmuv3")) {
2621         vms->iommu = VIRT_IOMMU_SMMUV3;
2622     } else if (!strcmp(value, "none")) {
2623         vms->iommu = VIRT_IOMMU_NONE;
2624     } else {
2625         error_setg(errp, "Invalid iommu value");
2626         error_append_hint(errp, "Valid values are none, smmuv3.\n");
2627     }
2628 }
2629 
2630 static bool virt_get_default_bus_bypass_iommu(Object *obj, Error **errp)
2631 {
2632     VirtMachineState *vms = VIRT_MACHINE(obj);
2633 
2634     return vms->default_bus_bypass_iommu;
2635 }
2636 
2637 static void virt_set_default_bus_bypass_iommu(Object *obj, bool value,
2638                                               Error **errp)
2639 {
2640     VirtMachineState *vms = VIRT_MACHINE(obj);
2641 
2642     vms->default_bus_bypass_iommu = value;
2643 }
2644 
2645 static CpuInstanceProperties
2646 virt_cpu_index_to_props(MachineState *ms, unsigned cpu_index)
2647 {
2648     MachineClass *mc = MACHINE_GET_CLASS(ms);
2649     const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms);
2650 
2651     assert(cpu_index < possible_cpus->len);
2652     return possible_cpus->cpus[cpu_index].props;
2653 }
2654 
2655 static int64_t virt_get_default_cpu_node_id(const MachineState *ms, int idx)
2656 {
2657     int64_t socket_id = ms->possible_cpus->cpus[idx].props.socket_id;
2658 
2659     return socket_id % ms->numa_state->num_nodes;
2660 }
2661 
2662 static const CPUArchIdList *virt_possible_cpu_arch_ids(MachineState *ms)
2663 {
2664     int n;
2665     unsigned int max_cpus = ms->smp.max_cpus;
2666     VirtMachineState *vms = VIRT_MACHINE(ms);
2667     MachineClass *mc = MACHINE_GET_CLASS(vms);
2668 
2669     if (ms->possible_cpus) {
2670         assert(ms->possible_cpus->len == max_cpus);
2671         return ms->possible_cpus;
2672     }
2673 
2674     ms->possible_cpus = g_malloc0(sizeof(CPUArchIdList) +
2675                                   sizeof(CPUArchId) * max_cpus);
2676     ms->possible_cpus->len = max_cpus;
2677     for (n = 0; n < ms->possible_cpus->len; n++) {
2678         ms->possible_cpus->cpus[n].type = ms->cpu_type;
2679         ms->possible_cpus->cpus[n].arch_id =
2680             virt_cpu_mp_affinity(vms, n);
2681 
2682         assert(!mc->smp_props.dies_supported);
2683         ms->possible_cpus->cpus[n].props.has_socket_id = true;
2684         ms->possible_cpus->cpus[n].props.socket_id =
2685             n / (ms->smp.clusters * ms->smp.cores * ms->smp.threads);
2686         ms->possible_cpus->cpus[n].props.has_cluster_id = true;
2687         ms->possible_cpus->cpus[n].props.cluster_id =
2688             (n / (ms->smp.cores * ms->smp.threads)) % ms->smp.clusters;
2689         ms->possible_cpus->cpus[n].props.has_core_id = true;
2690         ms->possible_cpus->cpus[n].props.core_id =
2691             (n / ms->smp.threads) % ms->smp.cores;
2692         ms->possible_cpus->cpus[n].props.has_thread_id = true;
2693         ms->possible_cpus->cpus[n].props.thread_id =
2694             n % ms->smp.threads;
2695     }
2696     return ms->possible_cpus;
2697 }
2698 
2699 static void virt_memory_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
2700                                  Error **errp)
2701 {
2702     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2703     const MachineState *ms = MACHINE(hotplug_dev);
2704     const bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2705 
2706     if (!vms->acpi_dev) {
2707         error_setg(errp,
2708                    "memory hotplug is not enabled: missing acpi-ged device");
2709         return;
2710     }
2711 
2712     if (vms->mte) {
2713         error_setg(errp, "memory hotplug is not enabled: MTE is enabled");
2714         return;
2715     }
2716 
2717     if (is_nvdimm && !ms->nvdimms_state->is_enabled) {
2718         error_setg(errp, "nvdimm is not enabled: add 'nvdimm=on' to '-M'");
2719         return;
2720     }
2721 
2722     pc_dimm_pre_plug(PC_DIMM(dev), MACHINE(hotplug_dev), NULL, errp);
2723 }
2724 
2725 static void virt_memory_plug(HotplugHandler *hotplug_dev,
2726                              DeviceState *dev, Error **errp)
2727 {
2728     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2729     MachineState *ms = MACHINE(hotplug_dev);
2730     bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2731 
2732     pc_dimm_plug(PC_DIMM(dev), MACHINE(vms));
2733 
2734     if (is_nvdimm) {
2735         nvdimm_plug(ms->nvdimms_state);
2736     }
2737 
2738     hotplug_handler_plug(HOTPLUG_HANDLER(vms->acpi_dev),
2739                          dev, &error_abort);
2740 }
2741 
2742 static void virt_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev,
2743                                             DeviceState *dev, Error **errp)
2744 {
2745     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2746 
2747     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2748         virt_memory_pre_plug(hotplug_dev, dev, errp);
2749     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI)) {
2750         virtio_md_pci_pre_plug(VIRTIO_MD_PCI(dev), MACHINE(hotplug_dev), errp);
2751     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2752         hwaddr db_start = 0, db_end = 0;
2753         QList *reserved_regions;
2754         char *resv_prop_str;
2755 
2756         if (vms->iommu != VIRT_IOMMU_NONE) {
2757             error_setg(errp, "virt machine does not support multiple IOMMUs");
2758             return;
2759         }
2760 
2761         switch (vms->msi_controller) {
2762         case VIRT_MSI_CTRL_NONE:
2763             return;
2764         case VIRT_MSI_CTRL_ITS:
2765             /* GITS_TRANSLATER page */
2766             db_start = base_memmap[VIRT_GIC_ITS].base + 0x10000;
2767             db_end = base_memmap[VIRT_GIC_ITS].base +
2768                      base_memmap[VIRT_GIC_ITS].size - 1;
2769             break;
2770         case VIRT_MSI_CTRL_GICV2M:
2771             /* MSI_SETSPI_NS page */
2772             db_start = base_memmap[VIRT_GIC_V2M].base;
2773             db_end = db_start + base_memmap[VIRT_GIC_V2M].size - 1;
2774             break;
2775         }
2776         resv_prop_str = g_strdup_printf("0x%"PRIx64":0x%"PRIx64":%u",
2777                                         db_start, db_end,
2778                                         VIRTIO_IOMMU_RESV_MEM_T_MSI);
2779 
2780         reserved_regions = qlist_new();
2781         qlist_append_str(reserved_regions, resv_prop_str);
2782         qdev_prop_set_array(dev, "reserved-regions", reserved_regions);
2783         g_free(resv_prop_str);
2784     }
2785 }
2786 
2787 static void virt_machine_device_plug_cb(HotplugHandler *hotplug_dev,
2788                                         DeviceState *dev, Error **errp)
2789 {
2790     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2791 
2792     if (vms->platform_bus_dev) {
2793         MachineClass *mc = MACHINE_GET_CLASS(vms);
2794 
2795         if (device_is_dynamic_sysbus(mc, dev)) {
2796             platform_bus_link_device(PLATFORM_BUS_DEVICE(vms->platform_bus_dev),
2797                                      SYS_BUS_DEVICE(dev));
2798         }
2799     }
2800 
2801     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2802         virt_memory_plug(hotplug_dev, dev, errp);
2803     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI)) {
2804         virtio_md_pci_plug(VIRTIO_MD_PCI(dev), MACHINE(hotplug_dev), errp);
2805     }
2806 
2807     if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2808         PCIDevice *pdev = PCI_DEVICE(dev);
2809 
2810         vms->iommu = VIRT_IOMMU_VIRTIO;
2811         vms->virtio_iommu_bdf = pci_get_bdf(pdev);
2812         create_virtio_iommu_dt_bindings(vms);
2813     }
2814 }
2815 
2816 static void virt_dimm_unplug_request(HotplugHandler *hotplug_dev,
2817                                      DeviceState *dev, Error **errp)
2818 {
2819     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
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 (object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM)) {
2828         error_setg(errp, "nvdimm device hot unplug is not supported yet.");
2829         return;
2830     }
2831 
2832     hotplug_handler_unplug_request(HOTPLUG_HANDLER(vms->acpi_dev), dev,
2833                                    errp);
2834 }
2835 
2836 static void virt_dimm_unplug(HotplugHandler *hotplug_dev,
2837                              DeviceState *dev, Error **errp)
2838 {
2839     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2840     Error *local_err = NULL;
2841 
2842     hotplug_handler_unplug(HOTPLUG_HANDLER(vms->acpi_dev), dev, &local_err);
2843     if (local_err) {
2844         goto out;
2845     }
2846 
2847     pc_dimm_unplug(PC_DIMM(dev), MACHINE(vms));
2848     qdev_unrealize(dev);
2849 
2850 out:
2851     error_propagate(errp, local_err);
2852 }
2853 
2854 static void virt_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
2855                                           DeviceState *dev, Error **errp)
2856 {
2857     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2858         virt_dimm_unplug_request(hotplug_dev, dev, errp);
2859     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI)) {
2860         virtio_md_pci_unplug_request(VIRTIO_MD_PCI(dev), MACHINE(hotplug_dev),
2861                                      errp);
2862     } else {
2863         error_setg(errp, "device unplug request for unsupported device"
2864                    " type: %s", object_get_typename(OBJECT(dev)));
2865     }
2866 }
2867 
2868 static void virt_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
2869                                           DeviceState *dev, Error **errp)
2870 {
2871     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2872         virt_dimm_unplug(hotplug_dev, dev, errp);
2873     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI)) {
2874         virtio_md_pci_unplug(VIRTIO_MD_PCI(dev), MACHINE(hotplug_dev), errp);
2875     } else {
2876         error_setg(errp, "virt: device unplug for unsupported device"
2877                    " type: %s", object_get_typename(OBJECT(dev)));
2878     }
2879 }
2880 
2881 static HotplugHandler *virt_machine_get_hotplug_handler(MachineState *machine,
2882                                                         DeviceState *dev)
2883 {
2884     MachineClass *mc = MACHINE_GET_CLASS(machine);
2885 
2886     if (device_is_dynamic_sysbus(mc, dev) ||
2887         object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) ||
2888         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI) ||
2889         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2890         return HOTPLUG_HANDLER(machine);
2891     }
2892     return NULL;
2893 }
2894 
2895 /*
2896  * for arm64 kvm_type [7-0] encodes the requested number of bits
2897  * in the IPA address space
2898  */
2899 static int virt_kvm_type(MachineState *ms, const char *type_str)
2900 {
2901     VirtMachineState *vms = VIRT_MACHINE(ms);
2902     int max_vm_pa_size, requested_pa_size;
2903     bool fixed_ipa;
2904 
2905     max_vm_pa_size = kvm_arm_get_max_vm_ipa_size(ms, &fixed_ipa);
2906 
2907     /* we freeze the memory map to compute the highest gpa */
2908     virt_set_memmap(vms, max_vm_pa_size);
2909 
2910     requested_pa_size = 64 - clz64(vms->highest_gpa);
2911 
2912     /*
2913      * KVM requires the IPA size to be at least 32 bits.
2914      */
2915     if (requested_pa_size < 32) {
2916         requested_pa_size = 32;
2917     }
2918 
2919     if (requested_pa_size > max_vm_pa_size) {
2920         error_report("-m and ,maxmem option values "
2921                      "require an IPA range (%d bits) larger than "
2922                      "the one supported by the host (%d bits)",
2923                      requested_pa_size, max_vm_pa_size);
2924         return -1;
2925     }
2926     /*
2927      * We return the requested PA log size, unless KVM only supports
2928      * the implicit legacy 40b IPA setting, in which case the kvm_type
2929      * must be 0.
2930      */
2931     return fixed_ipa ? 0 : requested_pa_size;
2932 }
2933 
2934 static void virt_machine_class_init(ObjectClass *oc, void *data)
2935 {
2936     MachineClass *mc = MACHINE_CLASS(oc);
2937     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
2938 
2939     mc->init = machvirt_init;
2940     /* Start with max_cpus set to 512, which is the maximum supported by KVM.
2941      * The value may be reduced later when we have more information about the
2942      * configuration of the particular instance.
2943      */
2944     mc->max_cpus = 512;
2945     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_CALXEDA_XGMAC);
2946     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_AMD_XGBE);
2947     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_RAMFB_DEVICE);
2948     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_PLATFORM);
2949 #ifdef CONFIG_TPM
2950     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_TPM_TIS_SYSBUS);
2951 #endif
2952     mc->block_default_type = IF_VIRTIO;
2953     mc->no_cdrom = 1;
2954     mc->pci_allow_0_address = true;
2955     /* We know we will never create a pre-ARMv7 CPU which needs 1K pages */
2956     mc->minimum_page_bits = 12;
2957     mc->possible_cpu_arch_ids = virt_possible_cpu_arch_ids;
2958     mc->cpu_index_to_instance_props = virt_cpu_index_to_props;
2959 #ifdef CONFIG_TCG
2960     mc->default_cpu_type = ARM_CPU_TYPE_NAME("cortex-a15");
2961 #else
2962     mc->default_cpu_type = ARM_CPU_TYPE_NAME("max");
2963 #endif
2964     mc->get_default_cpu_node_id = virt_get_default_cpu_node_id;
2965     mc->kvm_type = virt_kvm_type;
2966     assert(!mc->get_hotplug_handler);
2967     mc->get_hotplug_handler = virt_machine_get_hotplug_handler;
2968     hc->pre_plug = virt_machine_device_pre_plug_cb;
2969     hc->plug = virt_machine_device_plug_cb;
2970     hc->unplug_request = virt_machine_device_unplug_request_cb;
2971     hc->unplug = virt_machine_device_unplug_cb;
2972     mc->nvdimm_supported = true;
2973     mc->smp_props.clusters_supported = true;
2974     mc->auto_enable_numa_with_memhp = true;
2975     mc->auto_enable_numa_with_memdev = true;
2976     /* platform instead of architectural choice */
2977     mc->cpu_cluster_has_numa_boundary = true;
2978     mc->default_ram_id = "mach-virt.ram";
2979     mc->default_nic = "virtio-net-pci";
2980 
2981     object_class_property_add(oc, "acpi", "OnOffAuto",
2982         virt_get_acpi, virt_set_acpi,
2983         NULL, NULL);
2984     object_class_property_set_description(oc, "acpi",
2985         "Enable ACPI");
2986     object_class_property_add_bool(oc, "secure", virt_get_secure,
2987                                    virt_set_secure);
2988     object_class_property_set_description(oc, "secure",
2989                                                 "Set on/off to enable/disable the ARM "
2990                                                 "Security Extensions (TrustZone)");
2991 
2992     object_class_property_add_bool(oc, "virtualization", virt_get_virt,
2993                                    virt_set_virt);
2994     object_class_property_set_description(oc, "virtualization",
2995                                           "Set on/off to enable/disable emulating a "
2996                                           "guest CPU which implements the ARM "
2997                                           "Virtualization Extensions");
2998 
2999     object_class_property_add_bool(oc, "highmem", virt_get_highmem,
3000                                    virt_set_highmem);
3001     object_class_property_set_description(oc, "highmem",
3002                                           "Set on/off to enable/disable using "
3003                                           "physical address space above 32 bits");
3004 
3005     object_class_property_add_bool(oc, "compact-highmem",
3006                                    virt_get_compact_highmem,
3007                                    virt_set_compact_highmem);
3008     object_class_property_set_description(oc, "compact-highmem",
3009                                           "Set on/off to enable/disable compact "
3010                                           "layout for high memory regions");
3011 
3012     object_class_property_add_bool(oc, "highmem-redists",
3013                                    virt_get_highmem_redists,
3014                                    virt_set_highmem_redists);
3015     object_class_property_set_description(oc, "highmem-redists",
3016                                           "Set on/off to enable/disable high "
3017                                           "memory region for GICv3 or GICv4 "
3018                                           "redistributor");
3019 
3020     object_class_property_add_bool(oc, "highmem-ecam",
3021                                    virt_get_highmem_ecam,
3022                                    virt_set_highmem_ecam);
3023     object_class_property_set_description(oc, "highmem-ecam",
3024                                           "Set on/off to enable/disable high "
3025                                           "memory region for PCI ECAM");
3026 
3027     object_class_property_add_bool(oc, "highmem-mmio",
3028                                    virt_get_highmem_mmio,
3029                                    virt_set_highmem_mmio);
3030     object_class_property_set_description(oc, "highmem-mmio",
3031                                           "Set on/off to enable/disable high "
3032                                           "memory region for PCI MMIO");
3033 
3034     object_class_property_add_str(oc, "gic-version", virt_get_gic_version,
3035                                   virt_set_gic_version);
3036     object_class_property_set_description(oc, "gic-version",
3037                                           "Set GIC version. "
3038                                           "Valid values are 2, 3, 4, host and max");
3039 
3040     object_class_property_add_str(oc, "iommu", virt_get_iommu, virt_set_iommu);
3041     object_class_property_set_description(oc, "iommu",
3042                                           "Set the IOMMU type. "
3043                                           "Valid values are none and smmuv3");
3044 
3045     object_class_property_add_bool(oc, "default-bus-bypass-iommu",
3046                                    virt_get_default_bus_bypass_iommu,
3047                                    virt_set_default_bus_bypass_iommu);
3048     object_class_property_set_description(oc, "default-bus-bypass-iommu",
3049                                           "Set on/off to enable/disable "
3050                                           "bypass_iommu for default root bus");
3051 
3052     object_class_property_add_bool(oc, "ras", virt_get_ras,
3053                                    virt_set_ras);
3054     object_class_property_set_description(oc, "ras",
3055                                           "Set on/off to enable/disable reporting host memory errors "
3056                                           "to a KVM guest using ACPI and guest external abort exceptions");
3057 
3058     object_class_property_add_bool(oc, "mte", virt_get_mte, virt_set_mte);
3059     object_class_property_set_description(oc, "mte",
3060                                           "Set on/off to enable/disable emulating a "
3061                                           "guest CPU which implements the ARM "
3062                                           "Memory Tagging Extension");
3063 
3064     object_class_property_add_bool(oc, "its", virt_get_its,
3065                                    virt_set_its);
3066     object_class_property_set_description(oc, "its",
3067                                           "Set on/off to enable/disable "
3068                                           "ITS instantiation");
3069 
3070     object_class_property_add_bool(oc, "dtb-randomness",
3071                                    virt_get_dtb_randomness,
3072                                    virt_set_dtb_randomness);
3073     object_class_property_set_description(oc, "dtb-randomness",
3074                                           "Set off to disable passing random or "
3075                                           "non-deterministic dtb nodes to guest");
3076 
3077     object_class_property_add_bool(oc, "dtb-kaslr-seed",
3078                                    virt_get_dtb_randomness,
3079                                    virt_set_dtb_randomness);
3080     object_class_property_set_description(oc, "dtb-kaslr-seed",
3081                                           "Deprecated synonym of dtb-randomness");
3082 
3083     object_class_property_add_str(oc, "x-oem-id",
3084                                   virt_get_oem_id,
3085                                   virt_set_oem_id);
3086     object_class_property_set_description(oc, "x-oem-id",
3087                                           "Override the default value of field OEMID "
3088                                           "in ACPI table header."
3089                                           "The string may be up to 6 bytes in size");
3090 
3091 
3092     object_class_property_add_str(oc, "x-oem-table-id",
3093                                   virt_get_oem_table_id,
3094                                   virt_set_oem_table_id);
3095     object_class_property_set_description(oc, "x-oem-table-id",
3096                                           "Override the default value of field OEM Table ID "
3097                                           "in ACPI table header."
3098                                           "The string may be up to 8 bytes in size");
3099 
3100 }
3101 
3102 static void virt_instance_init(Object *obj)
3103 {
3104     VirtMachineState *vms = VIRT_MACHINE(obj);
3105     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
3106 
3107     /* EL3 is disabled by default on virt: this makes us consistent
3108      * between KVM and TCG for this board, and it also allows us to
3109      * boot UEFI blobs which assume no TrustZone support.
3110      */
3111     vms->secure = false;
3112 
3113     /* EL2 is also disabled by default, for similar reasons */
3114     vms->virt = false;
3115 
3116     /* High memory is enabled by default */
3117     vms->highmem = true;
3118     vms->highmem_compact = !vmc->no_highmem_compact;
3119     vms->gic_version = VIRT_GIC_VERSION_NOSEL;
3120 
3121     vms->highmem_ecam = !vmc->no_highmem_ecam;
3122     vms->highmem_mmio = true;
3123     vms->highmem_redists = true;
3124 
3125     if (vmc->no_its) {
3126         vms->its = false;
3127     } else {
3128         /* Default allows ITS instantiation */
3129         vms->its = true;
3130 
3131         if (vmc->no_tcg_its) {
3132             vms->tcg_its = false;
3133         } else {
3134             vms->tcg_its = true;
3135         }
3136     }
3137 
3138     /* Default disallows iommu instantiation */
3139     vms->iommu = VIRT_IOMMU_NONE;
3140 
3141     /* The default root bus is attached to iommu by default */
3142     vms->default_bus_bypass_iommu = false;
3143 
3144     /* Default disallows RAS instantiation */
3145     vms->ras = false;
3146 
3147     /* MTE is disabled by default.  */
3148     vms->mte = false;
3149 
3150     /* Supply kaslr-seed and rng-seed by default */
3151     vms->dtb_randomness = true;
3152 
3153     vms->irqmap = a15irqmap;
3154 
3155     virt_flash_create(vms);
3156 
3157     vms->oem_id = g_strndup(ACPI_BUILD_APPNAME6, 6);
3158     vms->oem_table_id = g_strndup(ACPI_BUILD_APPNAME8, 8);
3159 }
3160 
3161 static const TypeInfo virt_machine_info = {
3162     .name          = TYPE_VIRT_MACHINE,
3163     .parent        = TYPE_MACHINE,
3164     .abstract      = true,
3165     .instance_size = sizeof(VirtMachineState),
3166     .class_size    = sizeof(VirtMachineClass),
3167     .class_init    = virt_machine_class_init,
3168     .instance_init = virt_instance_init,
3169     .interfaces = (InterfaceInfo[]) {
3170          { TYPE_HOTPLUG_HANDLER },
3171          { }
3172     },
3173 };
3174 
3175 static void machvirt_machine_init(void)
3176 {
3177     type_register_static(&virt_machine_info);
3178 }
3179 type_init(machvirt_machine_init);
3180 
3181 static void virt_machine_8_2_options(MachineClass *mc)
3182 {
3183 }
3184 DEFINE_VIRT_MACHINE_AS_LATEST(8, 2)
3185 
3186 static void virt_machine_8_1_options(MachineClass *mc)
3187 {
3188     virt_machine_8_2_options(mc);
3189     compat_props_add(mc->compat_props, hw_compat_8_1, hw_compat_8_1_len);
3190 }
3191 DEFINE_VIRT_MACHINE(8, 1)
3192 
3193 static void virt_machine_8_0_options(MachineClass *mc)
3194 {
3195     virt_machine_8_1_options(mc);
3196     compat_props_add(mc->compat_props, hw_compat_8_0, hw_compat_8_0_len);
3197 }
3198 DEFINE_VIRT_MACHINE(8, 0)
3199 
3200 static void virt_machine_7_2_options(MachineClass *mc)
3201 {
3202     virt_machine_8_0_options(mc);
3203     compat_props_add(mc->compat_props, hw_compat_7_2, hw_compat_7_2_len);
3204 }
3205 DEFINE_VIRT_MACHINE(7, 2)
3206 
3207 static void virt_machine_7_1_options(MachineClass *mc)
3208 {
3209     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3210 
3211     virt_machine_7_2_options(mc);
3212     compat_props_add(mc->compat_props, hw_compat_7_1, hw_compat_7_1_len);
3213     /* Compact layout for high memory regions was introduced with 7.2 */
3214     vmc->no_highmem_compact = true;
3215 }
3216 DEFINE_VIRT_MACHINE(7, 1)
3217 
3218 static void virt_machine_7_0_options(MachineClass *mc)
3219 {
3220     virt_machine_7_1_options(mc);
3221     compat_props_add(mc->compat_props, hw_compat_7_0, hw_compat_7_0_len);
3222 }
3223 DEFINE_VIRT_MACHINE(7, 0)
3224 
3225 static void virt_machine_6_2_options(MachineClass *mc)
3226 {
3227     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3228 
3229     virt_machine_7_0_options(mc);
3230     compat_props_add(mc->compat_props, hw_compat_6_2, hw_compat_6_2_len);
3231     vmc->no_tcg_lpa2 = true;
3232 }
3233 DEFINE_VIRT_MACHINE(6, 2)
3234 
3235 static void virt_machine_6_1_options(MachineClass *mc)
3236 {
3237     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3238 
3239     virt_machine_6_2_options(mc);
3240     compat_props_add(mc->compat_props, hw_compat_6_1, hw_compat_6_1_len);
3241     mc->smp_props.prefer_sockets = true;
3242     vmc->no_cpu_topology = true;
3243 
3244     /* qemu ITS was introduced with 6.2 */
3245     vmc->no_tcg_its = true;
3246 }
3247 DEFINE_VIRT_MACHINE(6, 1)
3248 
3249 static void virt_machine_6_0_options(MachineClass *mc)
3250 {
3251     virt_machine_6_1_options(mc);
3252     compat_props_add(mc->compat_props, hw_compat_6_0, hw_compat_6_0_len);
3253 }
3254 DEFINE_VIRT_MACHINE(6, 0)
3255 
3256 static void virt_machine_5_2_options(MachineClass *mc)
3257 {
3258     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3259 
3260     virt_machine_6_0_options(mc);
3261     compat_props_add(mc->compat_props, hw_compat_5_2, hw_compat_5_2_len);
3262     vmc->no_secure_gpio = true;
3263 }
3264 DEFINE_VIRT_MACHINE(5, 2)
3265 
3266 static void virt_machine_5_1_options(MachineClass *mc)
3267 {
3268     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3269 
3270     virt_machine_5_2_options(mc);
3271     compat_props_add(mc->compat_props, hw_compat_5_1, hw_compat_5_1_len);
3272     vmc->no_kvm_steal_time = true;
3273 }
3274 DEFINE_VIRT_MACHINE(5, 1)
3275 
3276 static void virt_machine_5_0_options(MachineClass *mc)
3277 {
3278     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3279 
3280     virt_machine_5_1_options(mc);
3281     compat_props_add(mc->compat_props, hw_compat_5_0, hw_compat_5_0_len);
3282     mc->numa_mem_supported = true;
3283     vmc->acpi_expose_flash = true;
3284     mc->auto_enable_numa_with_memdev = false;
3285 }
3286 DEFINE_VIRT_MACHINE(5, 0)
3287 
3288 static void virt_machine_4_2_options(MachineClass *mc)
3289 {
3290     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3291 
3292     virt_machine_5_0_options(mc);
3293     compat_props_add(mc->compat_props, hw_compat_4_2, hw_compat_4_2_len);
3294     vmc->kvm_no_adjvtime = true;
3295 }
3296 DEFINE_VIRT_MACHINE(4, 2)
3297 
3298 static void virt_machine_4_1_options(MachineClass *mc)
3299 {
3300     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3301 
3302     virt_machine_4_2_options(mc);
3303     compat_props_add(mc->compat_props, hw_compat_4_1, hw_compat_4_1_len);
3304     vmc->no_ged = true;
3305     mc->auto_enable_numa_with_memhp = false;
3306 }
3307 DEFINE_VIRT_MACHINE(4, 1)
3308 
3309 static void virt_machine_4_0_options(MachineClass *mc)
3310 {
3311     virt_machine_4_1_options(mc);
3312     compat_props_add(mc->compat_props, hw_compat_4_0, hw_compat_4_0_len);
3313 }
3314 DEFINE_VIRT_MACHINE(4, 0)
3315 
3316 static void virt_machine_3_1_options(MachineClass *mc)
3317 {
3318     virt_machine_4_0_options(mc);
3319     compat_props_add(mc->compat_props, hw_compat_3_1, hw_compat_3_1_len);
3320 }
3321 DEFINE_VIRT_MACHINE(3, 1)
3322 
3323 static void virt_machine_3_0_options(MachineClass *mc)
3324 {
3325     virt_machine_3_1_options(mc);
3326     compat_props_add(mc->compat_props, hw_compat_3_0, hw_compat_3_0_len);
3327 }
3328 DEFINE_VIRT_MACHINE(3, 0)
3329 
3330 static void virt_machine_2_12_options(MachineClass *mc)
3331 {
3332     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3333 
3334     virt_machine_3_0_options(mc);
3335     compat_props_add(mc->compat_props, hw_compat_2_12, hw_compat_2_12_len);
3336     vmc->no_highmem_ecam = true;
3337     mc->max_cpus = 255;
3338 }
3339 DEFINE_VIRT_MACHINE(2, 12)
3340 
3341 static void virt_machine_2_11_options(MachineClass *mc)
3342 {
3343     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3344 
3345     virt_machine_2_12_options(mc);
3346     compat_props_add(mc->compat_props, hw_compat_2_11, hw_compat_2_11_len);
3347     vmc->smbios_old_sys_ver = true;
3348 }
3349 DEFINE_VIRT_MACHINE(2, 11)
3350 
3351 static void virt_machine_2_10_options(MachineClass *mc)
3352 {
3353     virt_machine_2_11_options(mc);
3354     compat_props_add(mc->compat_props, hw_compat_2_10, hw_compat_2_10_len);
3355     /* before 2.11 we never faulted accesses to bad addresses */
3356     mc->ignore_memory_transaction_failures = true;
3357 }
3358 DEFINE_VIRT_MACHINE(2, 10)
3359 
3360 static void virt_machine_2_9_options(MachineClass *mc)
3361 {
3362     virt_machine_2_10_options(mc);
3363     compat_props_add(mc->compat_props, hw_compat_2_9, hw_compat_2_9_len);
3364 }
3365 DEFINE_VIRT_MACHINE(2, 9)
3366 
3367 static void virt_machine_2_8_options(MachineClass *mc)
3368 {
3369     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3370 
3371     virt_machine_2_9_options(mc);
3372     compat_props_add(mc->compat_props, hw_compat_2_8, hw_compat_2_8_len);
3373     /* For 2.8 and earlier we falsely claimed in the DT that
3374      * our timers were edge-triggered, not level-triggered.
3375      */
3376     vmc->claim_edge_triggered_timers = true;
3377 }
3378 DEFINE_VIRT_MACHINE(2, 8)
3379 
3380 static void virt_machine_2_7_options(MachineClass *mc)
3381 {
3382     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3383 
3384     virt_machine_2_8_options(mc);
3385     compat_props_add(mc->compat_props, hw_compat_2_7, hw_compat_2_7_len);
3386     /* ITS was introduced with 2.8 */
3387     vmc->no_its = true;
3388     /* Stick with 1K pages for migration compatibility */
3389     mc->minimum_page_bits = 0;
3390 }
3391 DEFINE_VIRT_MACHINE(2, 7)
3392 
3393 static void virt_machine_2_6_options(MachineClass *mc)
3394 {
3395     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3396 
3397     virt_machine_2_7_options(mc);
3398     compat_props_add(mc->compat_props, hw_compat_2_6, hw_compat_2_6_len);
3399     vmc->disallow_affinity_adjustment = true;
3400     /* Disable PMU for 2.6 as PMU support was first introduced in 2.7 */
3401     vmc->no_pmu = true;
3402 }
3403 DEFINE_VIRT_MACHINE(2, 6)
3404