xref: /openbmc/qemu/hw/arm/virt.c (revision 36df75a0)
1 /*
2  * ARM mach-virt emulation
3  *
4  * Copyright (c) 2013 Linaro Limited
5  *
6  * This program is free software; you can redistribute it and/or modify it
7  * under the terms and conditions of the GNU General Public License,
8  * version 2 or later, as published by the Free Software Foundation.
9  *
10  * This program is distributed in the hope it will be useful, but WITHOUT
11  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
13  * more details.
14  *
15  * You should have received a copy of the GNU General Public License along with
16  * this program.  If not, see <http://www.gnu.org/licenses/>.
17  *
18  * Emulate a virtual board which works by passing Linux all the information
19  * it needs about what devices are present via the device tree.
20  * There are some restrictions about what we can do here:
21  *  + we can only present devices whose Linux drivers will work based
22  *    purely on the device tree with no platform data at all
23  *  + we want to present a very stripped-down minimalist platform,
24  *    both because this reduces the security attack surface from the guest
25  *    and also because it reduces our exposure to being broken when
26  *    the kernel updates its device tree bindings and requires further
27  *    information in a device binding that we aren't providing.
28  * This is essentially the same approach kvmtool uses.
29  */
30 
31 #include "qemu/osdep.h"
32 #include "qemu/datadir.h"
33 #include "qemu/units.h"
34 #include "qemu/option.h"
35 #include "monitor/qdev.h"
36 #include "hw/sysbus.h"
37 #include "hw/arm/boot.h"
38 #include "hw/arm/primecell.h"
39 #include "hw/arm/virt.h"
40 #include "hw/block/flash.h"
41 #include "hw/vfio/vfio-calxeda-xgmac.h"
42 #include "hw/vfio/vfio-amd-xgbe.h"
43 #include "hw/display/ramfb.h"
44 #include "net/net.h"
45 #include "sysemu/device_tree.h"
46 #include "sysemu/numa.h"
47 #include "sysemu/runstate.h"
48 #include "sysemu/tpm.h"
49 #include "sysemu/tcg.h"
50 #include "sysemu/kvm.h"
51 #include "sysemu/hvf.h"
52 #include "sysemu/qtest.h"
53 #include "hw/loader.h"
54 #include "qapi/error.h"
55 #include "qemu/bitops.h"
56 #include "qemu/error-report.h"
57 #include "qemu/module.h"
58 #include "hw/pci-host/gpex.h"
59 #include "hw/virtio/virtio-pci.h"
60 #include "hw/core/sysbus-fdt.h"
61 #include "hw/platform-bus.h"
62 #include "hw/qdev-properties.h"
63 #include "hw/arm/fdt.h"
64 #include "hw/intc/arm_gic.h"
65 #include "hw/intc/arm_gicv3_common.h"
66 #include "hw/intc/arm_gicv3_its_common.h"
67 #include "hw/irq.h"
68 #include "kvm_arm.h"
69 #include "hw/firmware/smbios.h"
70 #include "qapi/visitor.h"
71 #include "qapi/qapi-visit-common.h"
72 #include "standard-headers/linux/input.h"
73 #include "hw/arm/smmuv3.h"
74 #include "hw/acpi/acpi.h"
75 #include "target/arm/internals.h"
76 #include "hw/mem/pc-dimm.h"
77 #include "hw/mem/nvdimm.h"
78 #include "hw/acpi/generic_event_device.h"
79 #include "hw/virtio/virtio-md-pci.h"
80 #include "hw/virtio/virtio-iommu.h"
81 #include "hw/char/pl011.h"
82 #include "qemu/guest-random.h"
83 
84 #define DEFINE_VIRT_MACHINE_LATEST(major, minor, latest) \
85     static void virt_##major##_##minor##_class_init(ObjectClass *oc, \
86                                                     void *data) \
87     { \
88         MachineClass *mc = MACHINE_CLASS(oc); \
89         virt_machine_##major##_##minor##_options(mc); \
90         mc->desc = "QEMU " # major "." # minor " ARM Virtual Machine"; \
91         if (latest) { \
92             mc->alias = "virt"; \
93         } \
94     } \
95     static const TypeInfo machvirt_##major##_##minor##_info = { \
96         .name = MACHINE_TYPE_NAME("virt-" # major "." # minor), \
97         .parent = TYPE_VIRT_MACHINE, \
98         .class_init = virt_##major##_##minor##_class_init, \
99     }; \
100     static void machvirt_machine_##major##_##minor##_init(void) \
101     { \
102         type_register_static(&machvirt_##major##_##minor##_info); \
103     } \
104     type_init(machvirt_machine_##major##_##minor##_init);
105 
106 #define DEFINE_VIRT_MACHINE_AS_LATEST(major, minor) \
107     DEFINE_VIRT_MACHINE_LATEST(major, minor, true)
108 #define DEFINE_VIRT_MACHINE(major, minor) \
109     DEFINE_VIRT_MACHINE_LATEST(major, minor, false)
110 
111 
112 /* Number of external interrupt lines to configure the GIC with */
113 #define NUM_IRQS 256
114 
115 #define PLATFORM_BUS_NUM_IRQS 64
116 
117 /* Legacy RAM limit in GB (< version 4.0) */
118 #define LEGACY_RAMLIMIT_GB 255
119 #define LEGACY_RAMLIMIT_BYTES (LEGACY_RAMLIMIT_GB * GiB)
120 
121 /* Addresses and sizes of our components.
122  * 0..128MB is space for a flash device so we can run bootrom code such as UEFI.
123  * 128MB..256MB is used for miscellaneous device I/O.
124  * 256MB..1GB is reserved for possible future PCI support (ie where the
125  * PCI memory window will go if we add a PCI host controller).
126  * 1GB and up is RAM (which may happily spill over into the
127  * high memory region beyond 4GB).
128  * This represents a compromise between how much RAM can be given to
129  * a 32 bit VM and leaving space for expansion and in particular for PCI.
130  * Note that devices should generally be placed at multiples of 0x10000,
131  * to accommodate guests using 64K pages.
132  */
133 static const MemMapEntry base_memmap[] = {
134     /* Space up to 0x8000000 is reserved for a boot ROM */
135     [VIRT_FLASH] =              {          0, 0x08000000 },
136     [VIRT_CPUPERIPHS] =         { 0x08000000, 0x00020000 },
137     /* GIC distributor and CPU interfaces sit inside the CPU peripheral space */
138     [VIRT_GIC_DIST] =           { 0x08000000, 0x00010000 },
139     [VIRT_GIC_CPU] =            { 0x08010000, 0x00010000 },
140     [VIRT_GIC_V2M] =            { 0x08020000, 0x00001000 },
141     [VIRT_GIC_HYP] =            { 0x08030000, 0x00010000 },
142     [VIRT_GIC_VCPU] =           { 0x08040000, 0x00010000 },
143     /* The space in between here is reserved for GICv3 CPU/vCPU/HYP */
144     [VIRT_GIC_ITS] =            { 0x08080000, 0x00020000 },
145     /* This redistributor space allows up to 2*64kB*123 CPUs */
146     [VIRT_GIC_REDIST] =         { 0x080A0000, 0x00F60000 },
147     [VIRT_UART] =               { 0x09000000, 0x00001000 },
148     [VIRT_RTC] =                { 0x09010000, 0x00001000 },
149     [VIRT_FW_CFG] =             { 0x09020000, 0x00000018 },
150     [VIRT_GPIO] =               { 0x09030000, 0x00001000 },
151     [VIRT_SECURE_UART] =        { 0x09040000, 0x00001000 },
152     [VIRT_SMMU] =               { 0x09050000, 0x00020000 },
153     [VIRT_PCDIMM_ACPI] =        { 0x09070000, MEMORY_HOTPLUG_IO_LEN },
154     [VIRT_ACPI_GED] =           { 0x09080000, ACPI_GED_EVT_SEL_LEN },
155     [VIRT_NVDIMM_ACPI] =        { 0x09090000, NVDIMM_ACPI_IO_LEN},
156     [VIRT_PVTIME] =             { 0x090a0000, 0x00010000 },
157     [VIRT_SECURE_GPIO] =        { 0x090b0000, 0x00001000 },
158     [VIRT_MMIO] =               { 0x0a000000, 0x00000200 },
159     /* ...repeating for a total of NUM_VIRTIO_TRANSPORTS, each of that size */
160     [VIRT_PLATFORM_BUS] =       { 0x0c000000, 0x02000000 },
161     [VIRT_SECURE_MEM] =         { 0x0e000000, 0x01000000 },
162     [VIRT_PCIE_MMIO] =          { 0x10000000, 0x2eff0000 },
163     [VIRT_PCIE_PIO] =           { 0x3eff0000, 0x00010000 },
164     [VIRT_PCIE_ECAM] =          { 0x3f000000, 0x01000000 },
165     /* Actual RAM size depends on initial RAM and device memory settings */
166     [VIRT_MEM] =                { GiB, LEGACY_RAMLIMIT_BYTES },
167 };
168 
169 /*
170  * Highmem IO Regions: This memory map is floating, located after the RAM.
171  * Each MemMapEntry base (GPA) will be dynamically computed, depending on the
172  * top of the RAM, so that its base get the same alignment as the size,
173  * ie. a 512GiB entry will be aligned on a 512GiB boundary. If there is
174  * less than 256GiB of RAM, the floating area starts at the 256GiB mark.
175  * Note the extended_memmap is sized so that it eventually also includes the
176  * base_memmap entries (VIRT_HIGH_GIC_REDIST2 index is greater than the last
177  * index of base_memmap).
178  *
179  * The memory map for these Highmem IO Regions can be in legacy or compact
180  * layout, depending on 'compact-highmem' property. With legacy layout, the
181  * PA space for one specific region is always reserved, even if the region
182  * has been disabled or doesn't fit into the PA space. However, the PA space
183  * for the region won't be reserved in these circumstances with compact layout.
184  */
185 static MemMapEntry extended_memmap[] = {
186     /* Additional 64 MB redist region (can contain up to 512 redistributors) */
187     [VIRT_HIGH_GIC_REDIST2] =   { 0x0, 64 * MiB },
188     [VIRT_HIGH_PCIE_ECAM] =     { 0x0, 256 * MiB },
189     /* Second PCIe window */
190     [VIRT_HIGH_PCIE_MMIO] =     { 0x0, 512 * GiB },
191 };
192 
193 static const int a15irqmap[] = {
194     [VIRT_UART] = 1,
195     [VIRT_RTC] = 2,
196     [VIRT_PCIE] = 3, /* ... to 6 */
197     [VIRT_GPIO] = 7,
198     [VIRT_SECURE_UART] = 8,
199     [VIRT_ACPI_GED] = 9,
200     [VIRT_MMIO] = 16, /* ...to 16 + NUM_VIRTIO_TRANSPORTS - 1 */
201     [VIRT_GIC_V2M] = 48, /* ...to 48 + NUM_GICV2M_SPIS - 1 */
202     [VIRT_SMMU] = 74,    /* ...to 74 + NUM_SMMU_IRQS - 1 */
203     [VIRT_PLATFORM_BUS] = 112, /* ...to 112 + PLATFORM_BUS_NUM_IRQS -1 */
204 };
205 
206 static const char *valid_cpus[] = {
207 #ifdef CONFIG_TCG
208     ARM_CPU_TYPE_NAME("cortex-a7"),
209     ARM_CPU_TYPE_NAME("cortex-a15"),
210     ARM_CPU_TYPE_NAME("cortex-a35"),
211     ARM_CPU_TYPE_NAME("cortex-a55"),
212     ARM_CPU_TYPE_NAME("cortex-a72"),
213     ARM_CPU_TYPE_NAME("cortex-a76"),
214     ARM_CPU_TYPE_NAME("a64fx"),
215     ARM_CPU_TYPE_NAME("neoverse-n1"),
216     ARM_CPU_TYPE_NAME("neoverse-v1"),
217 #endif
218     ARM_CPU_TYPE_NAME("cortex-a53"),
219     ARM_CPU_TYPE_NAME("cortex-a57"),
220     ARM_CPU_TYPE_NAME("host"),
221     ARM_CPU_TYPE_NAME("max"),
222 };
223 
224 static bool cpu_type_valid(const char *cpu)
225 {
226     int i;
227 
228     for (i = 0; i < ARRAY_SIZE(valid_cpus); i++) {
229         if (strcmp(cpu, valid_cpus[i]) == 0) {
230             return true;
231         }
232     }
233     return false;
234 }
235 
236 static void create_randomness(MachineState *ms, const char *node)
237 {
238     struct {
239         uint64_t kaslr;
240         uint8_t rng[32];
241     } seed;
242 
243     if (qemu_guest_getrandom(&seed, sizeof(seed), NULL)) {
244         return;
245     }
246     qemu_fdt_setprop_u64(ms->fdt, node, "kaslr-seed", seed.kaslr);
247     qemu_fdt_setprop(ms->fdt, node, "rng-seed", seed.rng, sizeof(seed.rng));
248 }
249 
250 static void create_fdt(VirtMachineState *vms)
251 {
252     MachineState *ms = MACHINE(vms);
253     int nb_numa_nodes = ms->numa_state->num_nodes;
254     void *fdt = create_device_tree(&vms->fdt_size);
255 
256     if (!fdt) {
257         error_report("create_device_tree() failed");
258         exit(1);
259     }
260 
261     ms->fdt = fdt;
262 
263     /* Header */
264     qemu_fdt_setprop_string(fdt, "/", "compatible", "linux,dummy-virt");
265     qemu_fdt_setprop_cell(fdt, "/", "#address-cells", 0x2);
266     qemu_fdt_setprop_cell(fdt, "/", "#size-cells", 0x2);
267     qemu_fdt_setprop_string(fdt, "/", "model", "linux,dummy-virt");
268 
269     /* /chosen must exist for load_dtb to fill in necessary properties later */
270     qemu_fdt_add_subnode(fdt, "/chosen");
271     if (vms->dtb_randomness) {
272         create_randomness(ms, "/chosen");
273     }
274 
275     if (vms->secure) {
276         qemu_fdt_add_subnode(fdt, "/secure-chosen");
277         if (vms->dtb_randomness) {
278             create_randomness(ms, "/secure-chosen");
279         }
280     }
281 
282     /* Clock node, for the benefit of the UART. The kernel device tree
283      * binding documentation claims the PL011 node clock properties are
284      * optional but in practice if you omit them the kernel refuses to
285      * probe for the device.
286      */
287     vms->clock_phandle = qemu_fdt_alloc_phandle(fdt);
288     qemu_fdt_add_subnode(fdt, "/apb-pclk");
289     qemu_fdt_setprop_string(fdt, "/apb-pclk", "compatible", "fixed-clock");
290     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "#clock-cells", 0x0);
291     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "clock-frequency", 24000000);
292     qemu_fdt_setprop_string(fdt, "/apb-pclk", "clock-output-names",
293                                 "clk24mhz");
294     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "phandle", vms->clock_phandle);
295 
296     if (nb_numa_nodes > 0 && ms->numa_state->have_numa_distance) {
297         int size = nb_numa_nodes * nb_numa_nodes * 3 * sizeof(uint32_t);
298         uint32_t *matrix = g_malloc0(size);
299         int idx, i, j;
300 
301         for (i = 0; i < nb_numa_nodes; i++) {
302             for (j = 0; j < nb_numa_nodes; j++) {
303                 idx = (i * nb_numa_nodes + j) * 3;
304                 matrix[idx + 0] = cpu_to_be32(i);
305                 matrix[idx + 1] = cpu_to_be32(j);
306                 matrix[idx + 2] =
307                     cpu_to_be32(ms->numa_state->nodes[i].distance[j]);
308             }
309         }
310 
311         qemu_fdt_add_subnode(fdt, "/distance-map");
312         qemu_fdt_setprop_string(fdt, "/distance-map", "compatible",
313                                 "numa-distance-map-v1");
314         qemu_fdt_setprop(fdt, "/distance-map", "distance-matrix",
315                          matrix, size);
316         g_free(matrix);
317     }
318 }
319 
320 static void fdt_add_timer_nodes(const VirtMachineState *vms)
321 {
322     /* On real hardware these interrupts are level-triggered.
323      * On KVM they were edge-triggered before host kernel version 4.4,
324      * and level-triggered afterwards.
325      * On emulated QEMU they are level-triggered.
326      *
327      * Getting the DTB info about them wrong is awkward for some
328      * guest kernels:
329      *  pre-4.8 ignore the DT and leave the interrupt configured
330      *   with whatever the GIC reset value (or the bootloader) left it at
331      *  4.8 before rc6 honour the incorrect data by programming it back
332      *   into the GIC, causing problems
333      *  4.8rc6 and later ignore the DT and always write "level triggered"
334      *   into the GIC
335      *
336      * For backwards-compatibility, virt-2.8 and earlier will continue
337      * to say these are edge-triggered, but later machines will report
338      * the correct information.
339      */
340     ARMCPU *armcpu;
341     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
342     uint32_t irqflags = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
343     MachineState *ms = MACHINE(vms);
344 
345     if (vmc->claim_edge_triggered_timers) {
346         irqflags = GIC_FDT_IRQ_FLAGS_EDGE_LO_HI;
347     }
348 
349     if (vms->gic_version == VIRT_GIC_VERSION_2) {
350         irqflags = deposit32(irqflags, GIC_FDT_IRQ_PPI_CPU_START,
351                              GIC_FDT_IRQ_PPI_CPU_WIDTH,
352                              (1 << MACHINE(vms)->smp.cpus) - 1);
353     }
354 
355     qemu_fdt_add_subnode(ms->fdt, "/timer");
356 
357     armcpu = ARM_CPU(qemu_get_cpu(0));
358     if (arm_feature(&armcpu->env, ARM_FEATURE_V8)) {
359         const char compat[] = "arm,armv8-timer\0arm,armv7-timer";
360         qemu_fdt_setprop(ms->fdt, "/timer", "compatible",
361                          compat, sizeof(compat));
362     } else {
363         qemu_fdt_setprop_string(ms->fdt, "/timer", "compatible",
364                                 "arm,armv7-timer");
365     }
366     qemu_fdt_setprop(ms->fdt, "/timer", "always-on", NULL, 0);
367     qemu_fdt_setprop_cells(ms->fdt, "/timer", "interrupts",
368                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_S_EL1_IRQ, irqflags,
369                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_NS_EL1_IRQ, irqflags,
370                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_VIRT_IRQ, irqflags,
371                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_NS_EL2_IRQ, irqflags);
372 }
373 
374 static void fdt_add_cpu_nodes(const VirtMachineState *vms)
375 {
376     int cpu;
377     int addr_cells = 1;
378     const MachineState *ms = MACHINE(vms);
379     const VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
380     int smp_cpus = ms->smp.cpus;
381 
382     /*
383      * See Linux Documentation/devicetree/bindings/arm/cpus.yaml
384      * On ARM v8 64-bit systems value should be set to 2,
385      * that corresponds to the MPIDR_EL1 register size.
386      * If MPIDR_EL1[63:32] value is equal to 0 on all CPUs
387      * in the system, #address-cells can be set to 1, since
388      * MPIDR_EL1[63:32] bits are not used for CPUs
389      * identification.
390      *
391      * Here we actually don't know whether our system is 32- or 64-bit one.
392      * The simplest way to go is to examine affinity IDs of all our CPUs. If
393      * at least one of them has Aff3 populated, we set #address-cells to 2.
394      */
395     for (cpu = 0; cpu < smp_cpus; cpu++) {
396         ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
397 
398         if (armcpu->mp_affinity & ARM_AFF3_MASK) {
399             addr_cells = 2;
400             break;
401         }
402     }
403 
404     qemu_fdt_add_subnode(ms->fdt, "/cpus");
405     qemu_fdt_setprop_cell(ms->fdt, "/cpus", "#address-cells", addr_cells);
406     qemu_fdt_setprop_cell(ms->fdt, "/cpus", "#size-cells", 0x0);
407 
408     for (cpu = smp_cpus - 1; cpu >= 0; cpu--) {
409         char *nodename = g_strdup_printf("/cpus/cpu@%d", cpu);
410         ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
411         CPUState *cs = CPU(armcpu);
412 
413         qemu_fdt_add_subnode(ms->fdt, nodename);
414         qemu_fdt_setprop_string(ms->fdt, nodename, "device_type", "cpu");
415         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
416                                     armcpu->dtb_compatible);
417 
418         if (vms->psci_conduit != QEMU_PSCI_CONDUIT_DISABLED && smp_cpus > 1) {
419             qemu_fdt_setprop_string(ms->fdt, nodename,
420                                         "enable-method", "psci");
421         }
422 
423         if (addr_cells == 2) {
424             qemu_fdt_setprop_u64(ms->fdt, nodename, "reg",
425                                  armcpu->mp_affinity);
426         } else {
427             qemu_fdt_setprop_cell(ms->fdt, nodename, "reg",
428                                   armcpu->mp_affinity);
429         }
430 
431         if (ms->possible_cpus->cpus[cs->cpu_index].props.has_node_id) {
432             qemu_fdt_setprop_cell(ms->fdt, nodename, "numa-node-id",
433                 ms->possible_cpus->cpus[cs->cpu_index].props.node_id);
434         }
435 
436         if (!vmc->no_cpu_topology) {
437             qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle",
438                                   qemu_fdt_alloc_phandle(ms->fdt));
439         }
440 
441         g_free(nodename);
442     }
443 
444     if (!vmc->no_cpu_topology) {
445         /*
446          * Add vCPU topology description through fdt node cpu-map.
447          *
448          * See Linux Documentation/devicetree/bindings/cpu/cpu-topology.txt
449          * In a SMP system, the hierarchy of CPUs can be defined through
450          * four entities that are used to describe the layout of CPUs in
451          * the system: socket/cluster/core/thread.
452          *
453          * A socket node represents the boundary of system physical package
454          * and its child nodes must be one or more cluster nodes. A system
455          * can contain several layers of clustering within a single physical
456          * package and cluster nodes can be contained in parent cluster nodes.
457          *
458          * Note: currently we only support one layer of clustering within
459          * each physical package.
460          */
461         qemu_fdt_add_subnode(ms->fdt, "/cpus/cpu-map");
462 
463         for (cpu = smp_cpus - 1; cpu >= 0; cpu--) {
464             char *cpu_path = g_strdup_printf("/cpus/cpu@%d", cpu);
465             char *map_path;
466 
467             if (ms->smp.threads > 1) {
468                 map_path = g_strdup_printf(
469                     "/cpus/cpu-map/socket%d/cluster%d/core%d/thread%d",
470                     cpu / (ms->smp.clusters * ms->smp.cores * ms->smp.threads),
471                     (cpu / (ms->smp.cores * ms->smp.threads)) % ms->smp.clusters,
472                     (cpu / ms->smp.threads) % ms->smp.cores,
473                     cpu % ms->smp.threads);
474             } else {
475                 map_path = g_strdup_printf(
476                     "/cpus/cpu-map/socket%d/cluster%d/core%d",
477                     cpu / (ms->smp.clusters * ms->smp.cores),
478                     (cpu / ms->smp.cores) % ms->smp.clusters,
479                     cpu % ms->smp.cores);
480             }
481             qemu_fdt_add_path(ms->fdt, map_path);
482             qemu_fdt_setprop_phandle(ms->fdt, map_path, "cpu", cpu_path);
483 
484             g_free(map_path);
485             g_free(cpu_path);
486         }
487     }
488 }
489 
490 static void fdt_add_its_gic_node(VirtMachineState *vms)
491 {
492     char *nodename;
493     MachineState *ms = MACHINE(vms);
494 
495     vms->msi_phandle = qemu_fdt_alloc_phandle(ms->fdt);
496     nodename = g_strdup_printf("/intc/its@%" PRIx64,
497                                vms->memmap[VIRT_GIC_ITS].base);
498     qemu_fdt_add_subnode(ms->fdt, nodename);
499     qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
500                             "arm,gic-v3-its");
501     qemu_fdt_setprop(ms->fdt, nodename, "msi-controller", NULL, 0);
502     qemu_fdt_setprop_cell(ms->fdt, nodename, "#msi-cells", 1);
503     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
504                                  2, vms->memmap[VIRT_GIC_ITS].base,
505                                  2, vms->memmap[VIRT_GIC_ITS].size);
506     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", vms->msi_phandle);
507     g_free(nodename);
508 }
509 
510 static void fdt_add_v2m_gic_node(VirtMachineState *vms)
511 {
512     MachineState *ms = MACHINE(vms);
513     char *nodename;
514 
515     nodename = g_strdup_printf("/intc/v2m@%" PRIx64,
516                                vms->memmap[VIRT_GIC_V2M].base);
517     vms->msi_phandle = qemu_fdt_alloc_phandle(ms->fdt);
518     qemu_fdt_add_subnode(ms->fdt, nodename);
519     qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
520                             "arm,gic-v2m-frame");
521     qemu_fdt_setprop(ms->fdt, nodename, "msi-controller", NULL, 0);
522     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
523                                  2, vms->memmap[VIRT_GIC_V2M].base,
524                                  2, vms->memmap[VIRT_GIC_V2M].size);
525     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", vms->msi_phandle);
526     g_free(nodename);
527 }
528 
529 static void fdt_add_gic_node(VirtMachineState *vms)
530 {
531     MachineState *ms = MACHINE(vms);
532     char *nodename;
533 
534     vms->gic_phandle = qemu_fdt_alloc_phandle(ms->fdt);
535     qemu_fdt_setprop_cell(ms->fdt, "/", "interrupt-parent", vms->gic_phandle);
536 
537     nodename = g_strdup_printf("/intc@%" PRIx64,
538                                vms->memmap[VIRT_GIC_DIST].base);
539     qemu_fdt_add_subnode(ms->fdt, nodename);
540     qemu_fdt_setprop_cell(ms->fdt, nodename, "#interrupt-cells", 3);
541     qemu_fdt_setprop(ms->fdt, nodename, "interrupt-controller", NULL, 0);
542     qemu_fdt_setprop_cell(ms->fdt, nodename, "#address-cells", 0x2);
543     qemu_fdt_setprop_cell(ms->fdt, nodename, "#size-cells", 0x2);
544     qemu_fdt_setprop(ms->fdt, nodename, "ranges", NULL, 0);
545     if (vms->gic_version != VIRT_GIC_VERSION_2) {
546         int nb_redist_regions = virt_gicv3_redist_region_count(vms);
547 
548         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
549                                 "arm,gic-v3");
550 
551         qemu_fdt_setprop_cell(ms->fdt, nodename,
552                               "#redistributor-regions", nb_redist_regions);
553 
554         if (nb_redist_regions == 1) {
555             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
556                                          2, vms->memmap[VIRT_GIC_DIST].base,
557                                          2, vms->memmap[VIRT_GIC_DIST].size,
558                                          2, vms->memmap[VIRT_GIC_REDIST].base,
559                                          2, vms->memmap[VIRT_GIC_REDIST].size);
560         } else {
561             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
562                                  2, vms->memmap[VIRT_GIC_DIST].base,
563                                  2, vms->memmap[VIRT_GIC_DIST].size,
564                                  2, vms->memmap[VIRT_GIC_REDIST].base,
565                                  2, vms->memmap[VIRT_GIC_REDIST].size,
566                                  2, vms->memmap[VIRT_HIGH_GIC_REDIST2].base,
567                                  2, vms->memmap[VIRT_HIGH_GIC_REDIST2].size);
568         }
569 
570         if (vms->virt) {
571             qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
572                                    GIC_FDT_IRQ_TYPE_PPI, ARCH_GIC_MAINT_IRQ,
573                                    GIC_FDT_IRQ_FLAGS_LEVEL_HI);
574         }
575     } else {
576         /* 'cortex-a15-gic' means 'GIC v2' */
577         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
578                                 "arm,cortex-a15-gic");
579         if (!vms->virt) {
580             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
581                                          2, vms->memmap[VIRT_GIC_DIST].base,
582                                          2, vms->memmap[VIRT_GIC_DIST].size,
583                                          2, vms->memmap[VIRT_GIC_CPU].base,
584                                          2, vms->memmap[VIRT_GIC_CPU].size);
585         } else {
586             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
587                                          2, vms->memmap[VIRT_GIC_DIST].base,
588                                          2, vms->memmap[VIRT_GIC_DIST].size,
589                                          2, vms->memmap[VIRT_GIC_CPU].base,
590                                          2, vms->memmap[VIRT_GIC_CPU].size,
591                                          2, vms->memmap[VIRT_GIC_HYP].base,
592                                          2, vms->memmap[VIRT_GIC_HYP].size,
593                                          2, vms->memmap[VIRT_GIC_VCPU].base,
594                                          2, vms->memmap[VIRT_GIC_VCPU].size);
595             qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
596                                    GIC_FDT_IRQ_TYPE_PPI, ARCH_GIC_MAINT_IRQ,
597                                    GIC_FDT_IRQ_FLAGS_LEVEL_HI);
598         }
599     }
600 
601     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", vms->gic_phandle);
602     g_free(nodename);
603 }
604 
605 static void fdt_add_pmu_nodes(const VirtMachineState *vms)
606 {
607     ARMCPU *armcpu = ARM_CPU(first_cpu);
608     uint32_t irqflags = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
609     MachineState *ms = MACHINE(vms);
610 
611     if (!arm_feature(&armcpu->env, ARM_FEATURE_PMU)) {
612         assert(!object_property_get_bool(OBJECT(armcpu), "pmu", NULL));
613         return;
614     }
615 
616     if (vms->gic_version == VIRT_GIC_VERSION_2) {
617         irqflags = deposit32(irqflags, GIC_FDT_IRQ_PPI_CPU_START,
618                              GIC_FDT_IRQ_PPI_CPU_WIDTH,
619                              (1 << MACHINE(vms)->smp.cpus) - 1);
620     }
621 
622     qemu_fdt_add_subnode(ms->fdt, "/pmu");
623     if (arm_feature(&armcpu->env, ARM_FEATURE_V8)) {
624         const char compat[] = "arm,armv8-pmuv3";
625         qemu_fdt_setprop(ms->fdt, "/pmu", "compatible",
626                          compat, sizeof(compat));
627         qemu_fdt_setprop_cells(ms->fdt, "/pmu", "interrupts",
628                                GIC_FDT_IRQ_TYPE_PPI, VIRTUAL_PMU_IRQ, irqflags);
629     }
630 }
631 
632 static inline DeviceState *create_acpi_ged(VirtMachineState *vms)
633 {
634     DeviceState *dev;
635     MachineState *ms = MACHINE(vms);
636     int irq = vms->irqmap[VIRT_ACPI_GED];
637     uint32_t event = ACPI_GED_PWR_DOWN_EVT;
638 
639     if (ms->ram_slots) {
640         event |= ACPI_GED_MEM_HOTPLUG_EVT;
641     }
642 
643     if (ms->nvdimms_state->is_enabled) {
644         event |= ACPI_GED_NVDIMM_HOTPLUG_EVT;
645     }
646 
647     dev = qdev_new(TYPE_ACPI_GED);
648     qdev_prop_set_uint32(dev, "ged-event", event);
649 
650     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_ACPI_GED].base);
651     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 1, vms->memmap[VIRT_PCDIMM_ACPI].base);
652     sysbus_connect_irq(SYS_BUS_DEVICE(dev), 0, qdev_get_gpio_in(vms->gic, irq));
653 
654     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
655 
656     return dev;
657 }
658 
659 static void create_its(VirtMachineState *vms)
660 {
661     const char *itsclass = its_class_name();
662     DeviceState *dev;
663 
664     if (!strcmp(itsclass, "arm-gicv3-its")) {
665         if (!vms->tcg_its) {
666             itsclass = NULL;
667         }
668     }
669 
670     if (!itsclass) {
671         /* Do nothing if not supported */
672         return;
673     }
674 
675     dev = qdev_new(itsclass);
676 
677     object_property_set_link(OBJECT(dev), "parent-gicv3", OBJECT(vms->gic),
678                              &error_abort);
679     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
680     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_ITS].base);
681 
682     fdt_add_its_gic_node(vms);
683     vms->msi_controller = VIRT_MSI_CTRL_ITS;
684 }
685 
686 static void create_v2m(VirtMachineState *vms)
687 {
688     int i;
689     int irq = vms->irqmap[VIRT_GIC_V2M];
690     DeviceState *dev;
691 
692     dev = qdev_new("arm-gicv2m");
693     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_V2M].base);
694     qdev_prop_set_uint32(dev, "base-spi", irq);
695     qdev_prop_set_uint32(dev, "num-spi", NUM_GICV2M_SPIS);
696     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
697 
698     for (i = 0; i < NUM_GICV2M_SPIS; i++) {
699         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
700                            qdev_get_gpio_in(vms->gic, irq + i));
701     }
702 
703     fdt_add_v2m_gic_node(vms);
704     vms->msi_controller = VIRT_MSI_CTRL_GICV2M;
705 }
706 
707 static void create_gic(VirtMachineState *vms, MemoryRegion *mem)
708 {
709     MachineState *ms = MACHINE(vms);
710     /* We create a standalone GIC */
711     SysBusDevice *gicbusdev;
712     const char *gictype;
713     int i;
714     unsigned int smp_cpus = ms->smp.cpus;
715     uint32_t nb_redist_regions = 0;
716     int revision;
717 
718     if (vms->gic_version == VIRT_GIC_VERSION_2) {
719         gictype = gic_class_name();
720     } else {
721         gictype = gicv3_class_name();
722     }
723 
724     switch (vms->gic_version) {
725     case VIRT_GIC_VERSION_2:
726         revision = 2;
727         break;
728     case VIRT_GIC_VERSION_3:
729         revision = 3;
730         break;
731     case VIRT_GIC_VERSION_4:
732         revision = 4;
733         break;
734     default:
735         g_assert_not_reached();
736     }
737     vms->gic = qdev_new(gictype);
738     qdev_prop_set_uint32(vms->gic, "revision", revision);
739     qdev_prop_set_uint32(vms->gic, "num-cpu", smp_cpus);
740     /* Note that the num-irq property counts both internal and external
741      * interrupts; there are always 32 of the former (mandated by GIC spec).
742      */
743     qdev_prop_set_uint32(vms->gic, "num-irq", NUM_IRQS + 32);
744     if (!kvm_irqchip_in_kernel()) {
745         qdev_prop_set_bit(vms->gic, "has-security-extensions", vms->secure);
746     }
747 
748     if (vms->gic_version != VIRT_GIC_VERSION_2) {
749         uint32_t redist0_capacity = virt_redist_capacity(vms, VIRT_GIC_REDIST);
750         uint32_t redist0_count = MIN(smp_cpus, redist0_capacity);
751 
752         nb_redist_regions = virt_gicv3_redist_region_count(vms);
753 
754         qdev_prop_set_uint32(vms->gic, "len-redist-region-count",
755                              nb_redist_regions);
756         qdev_prop_set_uint32(vms->gic, "redist-region-count[0]", redist0_count);
757 
758         if (!kvm_irqchip_in_kernel()) {
759             if (vms->tcg_its) {
760                 object_property_set_link(OBJECT(vms->gic), "sysmem",
761                                          OBJECT(mem), &error_fatal);
762                 qdev_prop_set_bit(vms->gic, "has-lpi", true);
763             }
764         }
765 
766         if (nb_redist_regions == 2) {
767             uint32_t redist1_capacity =
768                 virt_redist_capacity(vms, VIRT_HIGH_GIC_REDIST2);
769 
770             qdev_prop_set_uint32(vms->gic, "redist-region-count[1]",
771                 MIN(smp_cpus - redist0_count, redist1_capacity));
772         }
773     } else {
774         if (!kvm_irqchip_in_kernel()) {
775             qdev_prop_set_bit(vms->gic, "has-virtualization-extensions",
776                               vms->virt);
777         }
778     }
779     gicbusdev = SYS_BUS_DEVICE(vms->gic);
780     sysbus_realize_and_unref(gicbusdev, &error_fatal);
781     sysbus_mmio_map(gicbusdev, 0, vms->memmap[VIRT_GIC_DIST].base);
782     if (vms->gic_version != VIRT_GIC_VERSION_2) {
783         sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_REDIST].base);
784         if (nb_redist_regions == 2) {
785             sysbus_mmio_map(gicbusdev, 2,
786                             vms->memmap[VIRT_HIGH_GIC_REDIST2].base);
787         }
788     } else {
789         sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_CPU].base);
790         if (vms->virt) {
791             sysbus_mmio_map(gicbusdev, 2, vms->memmap[VIRT_GIC_HYP].base);
792             sysbus_mmio_map(gicbusdev, 3, vms->memmap[VIRT_GIC_VCPU].base);
793         }
794     }
795 
796     /* Wire the outputs from each CPU's generic timer and the GICv3
797      * maintenance interrupt signal to the appropriate GIC PPI inputs,
798      * and the GIC's IRQ/FIQ/VIRQ/VFIQ interrupt outputs to the CPU's inputs.
799      */
800     for (i = 0; i < smp_cpus; i++) {
801         DeviceState *cpudev = DEVICE(qemu_get_cpu(i));
802         int ppibase = NUM_IRQS + i * GIC_INTERNAL + GIC_NR_SGIS;
803         int irq;
804         /* Mapping from the output timer irq lines from the CPU to the
805          * GIC PPI inputs we use for the virt board.
806          */
807         const int timer_irq[] = {
808             [GTIMER_PHYS] = ARCH_TIMER_NS_EL1_IRQ,
809             [GTIMER_VIRT] = ARCH_TIMER_VIRT_IRQ,
810             [GTIMER_HYP]  = ARCH_TIMER_NS_EL2_IRQ,
811             [GTIMER_SEC]  = ARCH_TIMER_S_EL1_IRQ,
812         };
813 
814         for (irq = 0; irq < ARRAY_SIZE(timer_irq); irq++) {
815             qdev_connect_gpio_out(cpudev, irq,
816                                   qdev_get_gpio_in(vms->gic,
817                                                    ppibase + timer_irq[irq]));
818         }
819 
820         if (vms->gic_version != VIRT_GIC_VERSION_2) {
821             qemu_irq irq = qdev_get_gpio_in(vms->gic,
822                                             ppibase + ARCH_GIC_MAINT_IRQ);
823             qdev_connect_gpio_out_named(cpudev, "gicv3-maintenance-interrupt",
824                                         0, irq);
825         } else if (vms->virt) {
826             qemu_irq irq = qdev_get_gpio_in(vms->gic,
827                                             ppibase + ARCH_GIC_MAINT_IRQ);
828             sysbus_connect_irq(gicbusdev, i + 4 * smp_cpus, irq);
829         }
830 
831         qdev_connect_gpio_out_named(cpudev, "pmu-interrupt", 0,
832                                     qdev_get_gpio_in(vms->gic, ppibase
833                                                      + VIRTUAL_PMU_IRQ));
834 
835         sysbus_connect_irq(gicbusdev, i, qdev_get_gpio_in(cpudev, ARM_CPU_IRQ));
836         sysbus_connect_irq(gicbusdev, i + smp_cpus,
837                            qdev_get_gpio_in(cpudev, ARM_CPU_FIQ));
838         sysbus_connect_irq(gicbusdev, i + 2 * smp_cpus,
839                            qdev_get_gpio_in(cpudev, ARM_CPU_VIRQ));
840         sysbus_connect_irq(gicbusdev, i + 3 * smp_cpus,
841                            qdev_get_gpio_in(cpudev, ARM_CPU_VFIQ));
842     }
843 
844     fdt_add_gic_node(vms);
845 
846     if (vms->gic_version != VIRT_GIC_VERSION_2 && vms->its) {
847         create_its(vms);
848     } else if (vms->gic_version == VIRT_GIC_VERSION_2) {
849         create_v2m(vms);
850     }
851 }
852 
853 static void create_uart(const VirtMachineState *vms, int uart,
854                         MemoryRegion *mem, Chardev *chr)
855 {
856     char *nodename;
857     hwaddr base = vms->memmap[uart].base;
858     hwaddr size = vms->memmap[uart].size;
859     int irq = vms->irqmap[uart];
860     const char compat[] = "arm,pl011\0arm,primecell";
861     const char clocknames[] = "uartclk\0apb_pclk";
862     DeviceState *dev = qdev_new(TYPE_PL011);
863     SysBusDevice *s = SYS_BUS_DEVICE(dev);
864     MachineState *ms = MACHINE(vms);
865 
866     qdev_prop_set_chr(dev, "chardev", chr);
867     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
868     memory_region_add_subregion(mem, base,
869                                 sysbus_mmio_get_region(s, 0));
870     sysbus_connect_irq(s, 0, qdev_get_gpio_in(vms->gic, irq));
871 
872     nodename = g_strdup_printf("/pl011@%" PRIx64, base);
873     qemu_fdt_add_subnode(ms->fdt, nodename);
874     /* Note that we can't use setprop_string because of the embedded NUL */
875     qemu_fdt_setprop(ms->fdt, nodename, "compatible",
876                          compat, sizeof(compat));
877     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
878                                      2, base, 2, size);
879     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
880                                GIC_FDT_IRQ_TYPE_SPI, irq,
881                                GIC_FDT_IRQ_FLAGS_LEVEL_HI);
882     qemu_fdt_setprop_cells(ms->fdt, nodename, "clocks",
883                                vms->clock_phandle, vms->clock_phandle);
884     qemu_fdt_setprop(ms->fdt, nodename, "clock-names",
885                          clocknames, sizeof(clocknames));
886 
887     if (uart == VIRT_UART) {
888         qemu_fdt_setprop_string(ms->fdt, "/chosen", "stdout-path", nodename);
889     } else {
890         /* Mark as not usable by the normal world */
891         qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
892         qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
893 
894         qemu_fdt_setprop_string(ms->fdt, "/secure-chosen", "stdout-path",
895                                 nodename);
896     }
897 
898     g_free(nodename);
899 }
900 
901 static void create_rtc(const VirtMachineState *vms)
902 {
903     char *nodename;
904     hwaddr base = vms->memmap[VIRT_RTC].base;
905     hwaddr size = vms->memmap[VIRT_RTC].size;
906     int irq = vms->irqmap[VIRT_RTC];
907     const char compat[] = "arm,pl031\0arm,primecell";
908     MachineState *ms = MACHINE(vms);
909 
910     sysbus_create_simple("pl031", base, qdev_get_gpio_in(vms->gic, irq));
911 
912     nodename = g_strdup_printf("/pl031@%" PRIx64, base);
913     qemu_fdt_add_subnode(ms->fdt, nodename);
914     qemu_fdt_setprop(ms->fdt, nodename, "compatible", compat, sizeof(compat));
915     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
916                                  2, base, 2, size);
917     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
918                            GIC_FDT_IRQ_TYPE_SPI, irq,
919                            GIC_FDT_IRQ_FLAGS_LEVEL_HI);
920     qemu_fdt_setprop_cell(ms->fdt, nodename, "clocks", vms->clock_phandle);
921     qemu_fdt_setprop_string(ms->fdt, nodename, "clock-names", "apb_pclk");
922     g_free(nodename);
923 }
924 
925 static DeviceState *gpio_key_dev;
926 static void virt_powerdown_req(Notifier *n, void *opaque)
927 {
928     VirtMachineState *s = container_of(n, VirtMachineState, powerdown_notifier);
929 
930     if (s->acpi_dev) {
931         acpi_send_event(s->acpi_dev, ACPI_POWER_DOWN_STATUS);
932     } else {
933         /* use gpio Pin 3 for power button event */
934         qemu_set_irq(qdev_get_gpio_in(gpio_key_dev, 0), 1);
935     }
936 }
937 
938 static void create_gpio_keys(char *fdt, DeviceState *pl061_dev,
939                              uint32_t phandle)
940 {
941     gpio_key_dev = sysbus_create_simple("gpio-key", -1,
942                                         qdev_get_gpio_in(pl061_dev, 3));
943 
944     qemu_fdt_add_subnode(fdt, "/gpio-keys");
945     qemu_fdt_setprop_string(fdt, "/gpio-keys", "compatible", "gpio-keys");
946 
947     qemu_fdt_add_subnode(fdt, "/gpio-keys/poweroff");
948     qemu_fdt_setprop_string(fdt, "/gpio-keys/poweroff",
949                             "label", "GPIO Key Poweroff");
950     qemu_fdt_setprop_cell(fdt, "/gpio-keys/poweroff", "linux,code",
951                           KEY_POWER);
952     qemu_fdt_setprop_cells(fdt, "/gpio-keys/poweroff",
953                            "gpios", phandle, 3, 0);
954 }
955 
956 #define SECURE_GPIO_POWEROFF 0
957 #define SECURE_GPIO_RESET    1
958 
959 static void create_secure_gpio_pwr(char *fdt, DeviceState *pl061_dev,
960                                    uint32_t phandle)
961 {
962     DeviceState *gpio_pwr_dev;
963 
964     /* gpio-pwr */
965     gpio_pwr_dev = sysbus_create_simple("gpio-pwr", -1, NULL);
966 
967     /* connect secure pl061 to gpio-pwr */
968     qdev_connect_gpio_out(pl061_dev, SECURE_GPIO_RESET,
969                           qdev_get_gpio_in_named(gpio_pwr_dev, "reset", 0));
970     qdev_connect_gpio_out(pl061_dev, SECURE_GPIO_POWEROFF,
971                           qdev_get_gpio_in_named(gpio_pwr_dev, "shutdown", 0));
972 
973     qemu_fdt_add_subnode(fdt, "/gpio-poweroff");
974     qemu_fdt_setprop_string(fdt, "/gpio-poweroff", "compatible",
975                             "gpio-poweroff");
976     qemu_fdt_setprop_cells(fdt, "/gpio-poweroff",
977                            "gpios", phandle, SECURE_GPIO_POWEROFF, 0);
978     qemu_fdt_setprop_string(fdt, "/gpio-poweroff", "status", "disabled");
979     qemu_fdt_setprop_string(fdt, "/gpio-poweroff", "secure-status",
980                             "okay");
981 
982     qemu_fdt_add_subnode(fdt, "/gpio-restart");
983     qemu_fdt_setprop_string(fdt, "/gpio-restart", "compatible",
984                             "gpio-restart");
985     qemu_fdt_setprop_cells(fdt, "/gpio-restart",
986                            "gpios", phandle, SECURE_GPIO_RESET, 0);
987     qemu_fdt_setprop_string(fdt, "/gpio-restart", "status", "disabled");
988     qemu_fdt_setprop_string(fdt, "/gpio-restart", "secure-status",
989                             "okay");
990 }
991 
992 static void create_gpio_devices(const VirtMachineState *vms, int gpio,
993                                 MemoryRegion *mem)
994 {
995     char *nodename;
996     DeviceState *pl061_dev;
997     hwaddr base = vms->memmap[gpio].base;
998     hwaddr size = vms->memmap[gpio].size;
999     int irq = vms->irqmap[gpio];
1000     const char compat[] = "arm,pl061\0arm,primecell";
1001     SysBusDevice *s;
1002     MachineState *ms = MACHINE(vms);
1003 
1004     pl061_dev = qdev_new("pl061");
1005     /* Pull lines down to 0 if not driven by the PL061 */
1006     qdev_prop_set_uint32(pl061_dev, "pullups", 0);
1007     qdev_prop_set_uint32(pl061_dev, "pulldowns", 0xff);
1008     s = SYS_BUS_DEVICE(pl061_dev);
1009     sysbus_realize_and_unref(s, &error_fatal);
1010     memory_region_add_subregion(mem, base, sysbus_mmio_get_region(s, 0));
1011     sysbus_connect_irq(s, 0, qdev_get_gpio_in(vms->gic, irq));
1012 
1013     uint32_t phandle = qemu_fdt_alloc_phandle(ms->fdt);
1014     nodename = g_strdup_printf("/pl061@%" PRIx64, base);
1015     qemu_fdt_add_subnode(ms->fdt, nodename);
1016     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1017                                  2, base, 2, size);
1018     qemu_fdt_setprop(ms->fdt, nodename, "compatible", compat, sizeof(compat));
1019     qemu_fdt_setprop_cell(ms->fdt, nodename, "#gpio-cells", 2);
1020     qemu_fdt_setprop(ms->fdt, nodename, "gpio-controller", NULL, 0);
1021     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
1022                            GIC_FDT_IRQ_TYPE_SPI, irq,
1023                            GIC_FDT_IRQ_FLAGS_LEVEL_HI);
1024     qemu_fdt_setprop_cell(ms->fdt, nodename, "clocks", vms->clock_phandle);
1025     qemu_fdt_setprop_string(ms->fdt, nodename, "clock-names", "apb_pclk");
1026     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", phandle);
1027 
1028     if (gpio != VIRT_GPIO) {
1029         /* Mark as not usable by the normal world */
1030         qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
1031         qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
1032     }
1033     g_free(nodename);
1034 
1035     /* Child gpio devices */
1036     if (gpio == VIRT_GPIO) {
1037         create_gpio_keys(ms->fdt, pl061_dev, phandle);
1038     } else {
1039         create_secure_gpio_pwr(ms->fdt, pl061_dev, phandle);
1040     }
1041 }
1042 
1043 static void create_virtio_devices(const VirtMachineState *vms)
1044 {
1045     int i;
1046     hwaddr size = vms->memmap[VIRT_MMIO].size;
1047     MachineState *ms = MACHINE(vms);
1048 
1049     /* We create the transports in forwards order. Since qbus_realize()
1050      * prepends (not appends) new child buses, the incrementing loop below will
1051      * create a list of virtio-mmio buses with decreasing base addresses.
1052      *
1053      * When a -device option is processed from the command line,
1054      * qbus_find_recursive() picks the next free virtio-mmio bus in forwards
1055      * order. The upshot is that -device options in increasing command line
1056      * order are mapped to virtio-mmio buses with decreasing base addresses.
1057      *
1058      * When this code was originally written, that arrangement ensured that the
1059      * guest Linux kernel would give the lowest "name" (/dev/vda, eth0, etc) to
1060      * the first -device on the command line. (The end-to-end order is a
1061      * function of this loop, qbus_realize(), qbus_find_recursive(), and the
1062      * guest kernel's name-to-address assignment strategy.)
1063      *
1064      * Meanwhile, the kernel's traversal seems to have been reversed; see eg.
1065      * the message, if not necessarily the code, of commit 70161ff336.
1066      * Therefore the loop now establishes the inverse of the original intent.
1067      *
1068      * Unfortunately, we can't counteract the kernel change by reversing the
1069      * loop; it would break existing command lines.
1070      *
1071      * In any case, the kernel makes no guarantee about the stability of
1072      * enumeration order of virtio devices (as demonstrated by it changing
1073      * between kernel versions). For reliable and stable identification
1074      * of disks users must use UUIDs or similar mechanisms.
1075      */
1076     for (i = 0; i < NUM_VIRTIO_TRANSPORTS; i++) {
1077         int irq = vms->irqmap[VIRT_MMIO] + i;
1078         hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
1079 
1080         sysbus_create_simple("virtio-mmio", base,
1081                              qdev_get_gpio_in(vms->gic, irq));
1082     }
1083 
1084     /* We add dtb nodes in reverse order so that they appear in the finished
1085      * device tree lowest address first.
1086      *
1087      * Note that this mapping is independent of the loop above. The previous
1088      * loop influences virtio device to virtio transport assignment, whereas
1089      * this loop controls how virtio transports are laid out in the dtb.
1090      */
1091     for (i = NUM_VIRTIO_TRANSPORTS - 1; i >= 0; i--) {
1092         char *nodename;
1093         int irq = vms->irqmap[VIRT_MMIO] + i;
1094         hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
1095 
1096         nodename = g_strdup_printf("/virtio_mmio@%" PRIx64, base);
1097         qemu_fdt_add_subnode(ms->fdt, nodename);
1098         qemu_fdt_setprop_string(ms->fdt, nodename,
1099                                 "compatible", "virtio,mmio");
1100         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1101                                      2, base, 2, size);
1102         qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
1103                                GIC_FDT_IRQ_TYPE_SPI, irq,
1104                                GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
1105         qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1106         g_free(nodename);
1107     }
1108 }
1109 
1110 #define VIRT_FLASH_SECTOR_SIZE (256 * KiB)
1111 
1112 static PFlashCFI01 *virt_flash_create1(VirtMachineState *vms,
1113                                         const char *name,
1114                                         const char *alias_prop_name)
1115 {
1116     /*
1117      * Create a single flash device.  We use the same parameters as
1118      * the flash devices on the Versatile Express board.
1119      */
1120     DeviceState *dev = qdev_new(TYPE_PFLASH_CFI01);
1121 
1122     qdev_prop_set_uint64(dev, "sector-length", VIRT_FLASH_SECTOR_SIZE);
1123     qdev_prop_set_uint8(dev, "width", 4);
1124     qdev_prop_set_uint8(dev, "device-width", 2);
1125     qdev_prop_set_bit(dev, "big-endian", false);
1126     qdev_prop_set_uint16(dev, "id0", 0x89);
1127     qdev_prop_set_uint16(dev, "id1", 0x18);
1128     qdev_prop_set_uint16(dev, "id2", 0x00);
1129     qdev_prop_set_uint16(dev, "id3", 0x00);
1130     qdev_prop_set_string(dev, "name", name);
1131     object_property_add_child(OBJECT(vms), name, OBJECT(dev));
1132     object_property_add_alias(OBJECT(vms), alias_prop_name,
1133                               OBJECT(dev), "drive");
1134     return PFLASH_CFI01(dev);
1135 }
1136 
1137 static void virt_flash_create(VirtMachineState *vms)
1138 {
1139     vms->flash[0] = virt_flash_create1(vms, "virt.flash0", "pflash0");
1140     vms->flash[1] = virt_flash_create1(vms, "virt.flash1", "pflash1");
1141 }
1142 
1143 static void virt_flash_map1(PFlashCFI01 *flash,
1144                             hwaddr base, hwaddr size,
1145                             MemoryRegion *sysmem)
1146 {
1147     DeviceState *dev = DEVICE(flash);
1148 
1149     assert(QEMU_IS_ALIGNED(size, VIRT_FLASH_SECTOR_SIZE));
1150     assert(size / VIRT_FLASH_SECTOR_SIZE <= UINT32_MAX);
1151     qdev_prop_set_uint32(dev, "num-blocks", size / VIRT_FLASH_SECTOR_SIZE);
1152     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1153 
1154     memory_region_add_subregion(sysmem, base,
1155                                 sysbus_mmio_get_region(SYS_BUS_DEVICE(dev),
1156                                                        0));
1157 }
1158 
1159 static void virt_flash_map(VirtMachineState *vms,
1160                            MemoryRegion *sysmem,
1161                            MemoryRegion *secure_sysmem)
1162 {
1163     /*
1164      * Map two flash devices to fill the VIRT_FLASH space in the memmap.
1165      * sysmem is the system memory space. secure_sysmem is the secure view
1166      * of the system, and the first flash device should be made visible only
1167      * there. The second flash device is visible to both secure and nonsecure.
1168      * If sysmem == secure_sysmem this means there is no separate Secure
1169      * address space and both flash devices are generally visible.
1170      */
1171     hwaddr flashsize = vms->memmap[VIRT_FLASH].size / 2;
1172     hwaddr flashbase = vms->memmap[VIRT_FLASH].base;
1173 
1174     virt_flash_map1(vms->flash[0], flashbase, flashsize,
1175                     secure_sysmem);
1176     virt_flash_map1(vms->flash[1], flashbase + flashsize, flashsize,
1177                     sysmem);
1178 }
1179 
1180 static void virt_flash_fdt(VirtMachineState *vms,
1181                            MemoryRegion *sysmem,
1182                            MemoryRegion *secure_sysmem)
1183 {
1184     hwaddr flashsize = vms->memmap[VIRT_FLASH].size / 2;
1185     hwaddr flashbase = vms->memmap[VIRT_FLASH].base;
1186     MachineState *ms = MACHINE(vms);
1187     char *nodename;
1188 
1189     if (sysmem == secure_sysmem) {
1190         /* Report both flash devices as a single node in the DT */
1191         nodename = g_strdup_printf("/flash@%" PRIx64, flashbase);
1192         qemu_fdt_add_subnode(ms->fdt, nodename);
1193         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible", "cfi-flash");
1194         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1195                                      2, flashbase, 2, flashsize,
1196                                      2, flashbase + flashsize, 2, flashsize);
1197         qemu_fdt_setprop_cell(ms->fdt, nodename, "bank-width", 4);
1198         g_free(nodename);
1199     } else {
1200         /*
1201          * Report the devices as separate nodes so we can mark one as
1202          * only visible to the secure world.
1203          */
1204         nodename = g_strdup_printf("/secflash@%" PRIx64, flashbase);
1205         qemu_fdt_add_subnode(ms->fdt, nodename);
1206         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible", "cfi-flash");
1207         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1208                                      2, flashbase, 2, flashsize);
1209         qemu_fdt_setprop_cell(ms->fdt, nodename, "bank-width", 4);
1210         qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
1211         qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
1212         g_free(nodename);
1213 
1214         nodename = g_strdup_printf("/flash@%" PRIx64, flashbase + flashsize);
1215         qemu_fdt_add_subnode(ms->fdt, nodename);
1216         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible", "cfi-flash");
1217         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1218                                      2, flashbase + flashsize, 2, flashsize);
1219         qemu_fdt_setprop_cell(ms->fdt, nodename, "bank-width", 4);
1220         g_free(nodename);
1221     }
1222 }
1223 
1224 static bool virt_firmware_init(VirtMachineState *vms,
1225                                MemoryRegion *sysmem,
1226                                MemoryRegion *secure_sysmem)
1227 {
1228     int i;
1229     const char *bios_name;
1230     BlockBackend *pflash_blk0;
1231 
1232     /* Map legacy -drive if=pflash to machine properties */
1233     for (i = 0; i < ARRAY_SIZE(vms->flash); i++) {
1234         pflash_cfi01_legacy_drive(vms->flash[i],
1235                                   drive_get(IF_PFLASH, 0, i));
1236     }
1237 
1238     virt_flash_map(vms, sysmem, secure_sysmem);
1239 
1240     pflash_blk0 = pflash_cfi01_get_blk(vms->flash[0]);
1241 
1242     bios_name = MACHINE(vms)->firmware;
1243     if (bios_name) {
1244         char *fname;
1245         MemoryRegion *mr;
1246         int image_size;
1247 
1248         if (pflash_blk0) {
1249             error_report("The contents of the first flash device may be "
1250                          "specified with -bios or with -drive if=pflash... "
1251                          "but you cannot use both options at once");
1252             exit(1);
1253         }
1254 
1255         /* Fall back to -bios */
1256 
1257         fname = qemu_find_file(QEMU_FILE_TYPE_BIOS, bios_name);
1258         if (!fname) {
1259             error_report("Could not find ROM image '%s'", bios_name);
1260             exit(1);
1261         }
1262         mr = sysbus_mmio_get_region(SYS_BUS_DEVICE(vms->flash[0]), 0);
1263         image_size = load_image_mr(fname, mr);
1264         g_free(fname);
1265         if (image_size < 0) {
1266             error_report("Could not load ROM image '%s'", bios_name);
1267             exit(1);
1268         }
1269     }
1270 
1271     return pflash_blk0 || bios_name;
1272 }
1273 
1274 static FWCfgState *create_fw_cfg(const VirtMachineState *vms, AddressSpace *as)
1275 {
1276     MachineState *ms = MACHINE(vms);
1277     hwaddr base = vms->memmap[VIRT_FW_CFG].base;
1278     hwaddr size = vms->memmap[VIRT_FW_CFG].size;
1279     FWCfgState *fw_cfg;
1280     char *nodename;
1281 
1282     fw_cfg = fw_cfg_init_mem_wide(base + 8, base, 8, base + 16, as);
1283     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, (uint16_t)ms->smp.cpus);
1284 
1285     nodename = g_strdup_printf("/fw-cfg@%" PRIx64, base);
1286     qemu_fdt_add_subnode(ms->fdt, nodename);
1287     qemu_fdt_setprop_string(ms->fdt, nodename,
1288                             "compatible", "qemu,fw-cfg-mmio");
1289     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1290                                  2, base, 2, size);
1291     qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1292     g_free(nodename);
1293     return fw_cfg;
1294 }
1295 
1296 static void create_pcie_irq_map(const MachineState *ms,
1297                                 uint32_t gic_phandle,
1298                                 int first_irq, const char *nodename)
1299 {
1300     int devfn, pin;
1301     uint32_t full_irq_map[4 * 4 * 10] = { 0 };
1302     uint32_t *irq_map = full_irq_map;
1303 
1304     for (devfn = 0; devfn <= 0x18; devfn += 0x8) {
1305         for (pin = 0; pin < 4; pin++) {
1306             int irq_type = GIC_FDT_IRQ_TYPE_SPI;
1307             int irq_nr = first_irq + ((pin + PCI_SLOT(devfn)) % PCI_NUM_PINS);
1308             int irq_level = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
1309             int i;
1310 
1311             uint32_t map[] = {
1312                 devfn << 8, 0, 0,                           /* devfn */
1313                 pin + 1,                                    /* PCI pin */
1314                 gic_phandle, 0, 0, irq_type, irq_nr, irq_level }; /* GIC irq */
1315 
1316             /* Convert map to big endian */
1317             for (i = 0; i < 10; i++) {
1318                 irq_map[i] = cpu_to_be32(map[i]);
1319             }
1320             irq_map += 10;
1321         }
1322     }
1323 
1324     qemu_fdt_setprop(ms->fdt, nodename, "interrupt-map",
1325                      full_irq_map, sizeof(full_irq_map));
1326 
1327     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupt-map-mask",
1328                            cpu_to_be16(PCI_DEVFN(3, 0)), /* Slot 3 */
1329                            0, 0,
1330                            0x7           /* PCI irq */);
1331 }
1332 
1333 static void create_smmu(const VirtMachineState *vms,
1334                         PCIBus *bus)
1335 {
1336     char *node;
1337     const char compat[] = "arm,smmu-v3";
1338     int irq =  vms->irqmap[VIRT_SMMU];
1339     int i;
1340     hwaddr base = vms->memmap[VIRT_SMMU].base;
1341     hwaddr size = vms->memmap[VIRT_SMMU].size;
1342     const char irq_names[] = "eventq\0priq\0cmdq-sync\0gerror";
1343     DeviceState *dev;
1344     MachineState *ms = MACHINE(vms);
1345 
1346     if (vms->iommu != VIRT_IOMMU_SMMUV3 || !vms->iommu_phandle) {
1347         return;
1348     }
1349 
1350     dev = qdev_new(TYPE_ARM_SMMUV3);
1351 
1352     object_property_set_link(OBJECT(dev), "primary-bus", OBJECT(bus),
1353                              &error_abort);
1354     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1355     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, base);
1356     for (i = 0; i < NUM_SMMU_IRQS; i++) {
1357         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
1358                            qdev_get_gpio_in(vms->gic, irq + i));
1359     }
1360 
1361     node = g_strdup_printf("/smmuv3@%" PRIx64, base);
1362     qemu_fdt_add_subnode(ms->fdt, node);
1363     qemu_fdt_setprop(ms->fdt, node, "compatible", compat, sizeof(compat));
1364     qemu_fdt_setprop_sized_cells(ms->fdt, node, "reg", 2, base, 2, size);
1365 
1366     qemu_fdt_setprop_cells(ms->fdt, node, "interrupts",
1367             GIC_FDT_IRQ_TYPE_SPI, irq    , GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1368             GIC_FDT_IRQ_TYPE_SPI, irq + 1, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1369             GIC_FDT_IRQ_TYPE_SPI, irq + 2, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1370             GIC_FDT_IRQ_TYPE_SPI, irq + 3, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
1371 
1372     qemu_fdt_setprop(ms->fdt, node, "interrupt-names", irq_names,
1373                      sizeof(irq_names));
1374 
1375     qemu_fdt_setprop(ms->fdt, node, "dma-coherent", NULL, 0);
1376 
1377     qemu_fdt_setprop_cell(ms->fdt, node, "#iommu-cells", 1);
1378 
1379     qemu_fdt_setprop_cell(ms->fdt, node, "phandle", vms->iommu_phandle);
1380     g_free(node);
1381 }
1382 
1383 static void create_virtio_iommu_dt_bindings(VirtMachineState *vms)
1384 {
1385     const char compat[] = "virtio,pci-iommu\0pci1af4,1057";
1386     uint16_t bdf = vms->virtio_iommu_bdf;
1387     MachineState *ms = MACHINE(vms);
1388     char *node;
1389 
1390     vms->iommu_phandle = qemu_fdt_alloc_phandle(ms->fdt);
1391 
1392     node = g_strdup_printf("%s/virtio_iommu@%x,%x", vms->pciehb_nodename,
1393                            PCI_SLOT(bdf), PCI_FUNC(bdf));
1394     qemu_fdt_add_subnode(ms->fdt, node);
1395     qemu_fdt_setprop(ms->fdt, node, "compatible", compat, sizeof(compat));
1396     qemu_fdt_setprop_sized_cells(ms->fdt, node, "reg",
1397                                  1, bdf << 8, 1, 0, 1, 0,
1398                                  1, 0, 1, 0);
1399 
1400     qemu_fdt_setprop_cell(ms->fdt, node, "#iommu-cells", 1);
1401     qemu_fdt_setprop_cell(ms->fdt, node, "phandle", vms->iommu_phandle);
1402     g_free(node);
1403 
1404     qemu_fdt_setprop_cells(ms->fdt, vms->pciehb_nodename, "iommu-map",
1405                            0x0, vms->iommu_phandle, 0x0, bdf,
1406                            bdf + 1, vms->iommu_phandle, bdf + 1, 0xffff - bdf);
1407 }
1408 
1409 static void create_pcie(VirtMachineState *vms)
1410 {
1411     hwaddr base_mmio = vms->memmap[VIRT_PCIE_MMIO].base;
1412     hwaddr size_mmio = vms->memmap[VIRT_PCIE_MMIO].size;
1413     hwaddr base_mmio_high = vms->memmap[VIRT_HIGH_PCIE_MMIO].base;
1414     hwaddr size_mmio_high = vms->memmap[VIRT_HIGH_PCIE_MMIO].size;
1415     hwaddr base_pio = vms->memmap[VIRT_PCIE_PIO].base;
1416     hwaddr size_pio = vms->memmap[VIRT_PCIE_PIO].size;
1417     hwaddr base_ecam, size_ecam;
1418     hwaddr base = base_mmio;
1419     int nr_pcie_buses;
1420     int irq = vms->irqmap[VIRT_PCIE];
1421     MemoryRegion *mmio_alias;
1422     MemoryRegion *mmio_reg;
1423     MemoryRegion *ecam_alias;
1424     MemoryRegion *ecam_reg;
1425     DeviceState *dev;
1426     char *nodename;
1427     int i, ecam_id;
1428     PCIHostState *pci;
1429     MachineState *ms = MACHINE(vms);
1430     MachineClass *mc = MACHINE_GET_CLASS(ms);
1431 
1432     dev = qdev_new(TYPE_GPEX_HOST);
1433     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1434 
1435     ecam_id = VIRT_ECAM_ID(vms->highmem_ecam);
1436     base_ecam = vms->memmap[ecam_id].base;
1437     size_ecam = vms->memmap[ecam_id].size;
1438     nr_pcie_buses = size_ecam / PCIE_MMCFG_SIZE_MIN;
1439     /* Map only the first size_ecam bytes of ECAM space */
1440     ecam_alias = g_new0(MemoryRegion, 1);
1441     ecam_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 0);
1442     memory_region_init_alias(ecam_alias, OBJECT(dev), "pcie-ecam",
1443                              ecam_reg, 0, size_ecam);
1444     memory_region_add_subregion(get_system_memory(), base_ecam, ecam_alias);
1445 
1446     /* Map the MMIO window into system address space so as to expose
1447      * the section of PCI MMIO space which starts at the same base address
1448      * (ie 1:1 mapping for that part of PCI MMIO space visible through
1449      * the window).
1450      */
1451     mmio_alias = g_new0(MemoryRegion, 1);
1452     mmio_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 1);
1453     memory_region_init_alias(mmio_alias, OBJECT(dev), "pcie-mmio",
1454                              mmio_reg, base_mmio, size_mmio);
1455     memory_region_add_subregion(get_system_memory(), base_mmio, mmio_alias);
1456 
1457     if (vms->highmem_mmio) {
1458         /* Map high MMIO space */
1459         MemoryRegion *high_mmio_alias = g_new0(MemoryRegion, 1);
1460 
1461         memory_region_init_alias(high_mmio_alias, OBJECT(dev), "pcie-mmio-high",
1462                                  mmio_reg, base_mmio_high, size_mmio_high);
1463         memory_region_add_subregion(get_system_memory(), base_mmio_high,
1464                                     high_mmio_alias);
1465     }
1466 
1467     /* Map IO port space */
1468     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 2, base_pio);
1469 
1470     for (i = 0; i < GPEX_NUM_IRQS; i++) {
1471         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
1472                            qdev_get_gpio_in(vms->gic, irq + i));
1473         gpex_set_irq_num(GPEX_HOST(dev), i, irq + i);
1474     }
1475 
1476     pci = PCI_HOST_BRIDGE(dev);
1477     pci->bypass_iommu = vms->default_bus_bypass_iommu;
1478     vms->bus = pci->bus;
1479     if (vms->bus) {
1480         for (i = 0; i < nb_nics; i++) {
1481             pci_nic_init_nofail(&nd_table[i], pci->bus, mc->default_nic, NULL);
1482         }
1483     }
1484 
1485     nodename = vms->pciehb_nodename = g_strdup_printf("/pcie@%" PRIx64, base);
1486     qemu_fdt_add_subnode(ms->fdt, nodename);
1487     qemu_fdt_setprop_string(ms->fdt, nodename,
1488                             "compatible", "pci-host-ecam-generic");
1489     qemu_fdt_setprop_string(ms->fdt, nodename, "device_type", "pci");
1490     qemu_fdt_setprop_cell(ms->fdt, nodename, "#address-cells", 3);
1491     qemu_fdt_setprop_cell(ms->fdt, nodename, "#size-cells", 2);
1492     qemu_fdt_setprop_cell(ms->fdt, nodename, "linux,pci-domain", 0);
1493     qemu_fdt_setprop_cells(ms->fdt, nodename, "bus-range", 0,
1494                            nr_pcie_buses - 1);
1495     qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1496 
1497     if (vms->msi_phandle) {
1498         qemu_fdt_setprop_cells(ms->fdt, nodename, "msi-map",
1499                                0, vms->msi_phandle, 0, 0x10000);
1500     }
1501 
1502     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1503                                  2, base_ecam, 2, size_ecam);
1504 
1505     if (vms->highmem_mmio) {
1506         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "ranges",
1507                                      1, FDT_PCI_RANGE_IOPORT, 2, 0,
1508                                      2, base_pio, 2, size_pio,
1509                                      1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1510                                      2, base_mmio, 2, size_mmio,
1511                                      1, FDT_PCI_RANGE_MMIO_64BIT,
1512                                      2, base_mmio_high,
1513                                      2, base_mmio_high, 2, size_mmio_high);
1514     } else {
1515         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "ranges",
1516                                      1, FDT_PCI_RANGE_IOPORT, 2, 0,
1517                                      2, base_pio, 2, size_pio,
1518                                      1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1519                                      2, base_mmio, 2, size_mmio);
1520     }
1521 
1522     qemu_fdt_setprop_cell(ms->fdt, nodename, "#interrupt-cells", 1);
1523     create_pcie_irq_map(ms, vms->gic_phandle, irq, nodename);
1524 
1525     if (vms->iommu) {
1526         vms->iommu_phandle = qemu_fdt_alloc_phandle(ms->fdt);
1527 
1528         switch (vms->iommu) {
1529         case VIRT_IOMMU_SMMUV3:
1530             create_smmu(vms, vms->bus);
1531             qemu_fdt_setprop_cells(ms->fdt, nodename, "iommu-map",
1532                                    0x0, vms->iommu_phandle, 0x0, 0x10000);
1533             break;
1534         default:
1535             g_assert_not_reached();
1536         }
1537     }
1538 }
1539 
1540 static void create_platform_bus(VirtMachineState *vms)
1541 {
1542     DeviceState *dev;
1543     SysBusDevice *s;
1544     int i;
1545     MemoryRegion *sysmem = get_system_memory();
1546 
1547     dev = qdev_new(TYPE_PLATFORM_BUS_DEVICE);
1548     dev->id = g_strdup(TYPE_PLATFORM_BUS_DEVICE);
1549     qdev_prop_set_uint32(dev, "num_irqs", PLATFORM_BUS_NUM_IRQS);
1550     qdev_prop_set_uint32(dev, "mmio_size", vms->memmap[VIRT_PLATFORM_BUS].size);
1551     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1552     vms->platform_bus_dev = dev;
1553 
1554     s = SYS_BUS_DEVICE(dev);
1555     for (i = 0; i < PLATFORM_BUS_NUM_IRQS; i++) {
1556         int irq = vms->irqmap[VIRT_PLATFORM_BUS] + i;
1557         sysbus_connect_irq(s, i, qdev_get_gpio_in(vms->gic, irq));
1558     }
1559 
1560     memory_region_add_subregion(sysmem,
1561                                 vms->memmap[VIRT_PLATFORM_BUS].base,
1562                                 sysbus_mmio_get_region(s, 0));
1563 }
1564 
1565 static void create_tag_ram(MemoryRegion *tag_sysmem,
1566                            hwaddr base, hwaddr size,
1567                            const char *name)
1568 {
1569     MemoryRegion *tagram = g_new(MemoryRegion, 1);
1570 
1571     memory_region_init_ram(tagram, NULL, name, size / 32, &error_fatal);
1572     memory_region_add_subregion(tag_sysmem, base / 32, tagram);
1573 }
1574 
1575 static void create_secure_ram(VirtMachineState *vms,
1576                               MemoryRegion *secure_sysmem,
1577                               MemoryRegion *secure_tag_sysmem)
1578 {
1579     MemoryRegion *secram = g_new(MemoryRegion, 1);
1580     char *nodename;
1581     hwaddr base = vms->memmap[VIRT_SECURE_MEM].base;
1582     hwaddr size = vms->memmap[VIRT_SECURE_MEM].size;
1583     MachineState *ms = MACHINE(vms);
1584 
1585     memory_region_init_ram(secram, NULL, "virt.secure-ram", size,
1586                            &error_fatal);
1587     memory_region_add_subregion(secure_sysmem, base, secram);
1588 
1589     nodename = g_strdup_printf("/secram@%" PRIx64, base);
1590     qemu_fdt_add_subnode(ms->fdt, nodename);
1591     qemu_fdt_setprop_string(ms->fdt, nodename, "device_type", "memory");
1592     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg", 2, base, 2, size);
1593     qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
1594     qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
1595 
1596     if (secure_tag_sysmem) {
1597         create_tag_ram(secure_tag_sysmem, base, size, "mach-virt.secure-tag");
1598     }
1599 
1600     g_free(nodename);
1601 }
1602 
1603 static void *machvirt_dtb(const struct arm_boot_info *binfo, int *fdt_size)
1604 {
1605     const VirtMachineState *board = container_of(binfo, VirtMachineState,
1606                                                  bootinfo);
1607     MachineState *ms = MACHINE(board);
1608 
1609 
1610     *fdt_size = board->fdt_size;
1611     return ms->fdt;
1612 }
1613 
1614 static void virt_build_smbios(VirtMachineState *vms)
1615 {
1616     MachineClass *mc = MACHINE_GET_CLASS(vms);
1617     MachineState *ms = MACHINE(vms);
1618     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1619     uint8_t *smbios_tables, *smbios_anchor;
1620     size_t smbios_tables_len, smbios_anchor_len;
1621     struct smbios_phys_mem_area mem_array;
1622     const char *product = "QEMU Virtual Machine";
1623 
1624     if (kvm_enabled()) {
1625         product = "KVM Virtual Machine";
1626     }
1627 
1628     smbios_set_defaults("QEMU", product,
1629                         vmc->smbios_old_sys_ver ? "1.0" : mc->name, false,
1630                         true, SMBIOS_ENTRY_POINT_TYPE_64);
1631 
1632     /* build the array of physical mem area from base_memmap */
1633     mem_array.address = vms->memmap[VIRT_MEM].base;
1634     mem_array.length = ms->ram_size;
1635 
1636     smbios_get_tables(ms, &mem_array, 1,
1637                       &smbios_tables, &smbios_tables_len,
1638                       &smbios_anchor, &smbios_anchor_len,
1639                       &error_fatal);
1640 
1641     if (smbios_anchor) {
1642         fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-tables",
1643                         smbios_tables, smbios_tables_len);
1644         fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-anchor",
1645                         smbios_anchor, smbios_anchor_len);
1646     }
1647 }
1648 
1649 static
1650 void virt_machine_done(Notifier *notifier, void *data)
1651 {
1652     VirtMachineState *vms = container_of(notifier, VirtMachineState,
1653                                          machine_done);
1654     MachineState *ms = MACHINE(vms);
1655     ARMCPU *cpu = ARM_CPU(first_cpu);
1656     struct arm_boot_info *info = &vms->bootinfo;
1657     AddressSpace *as = arm_boot_address_space(cpu, info);
1658 
1659     /*
1660      * If the user provided a dtb, we assume the dynamic sysbus nodes
1661      * already are integrated there. This corresponds to a use case where
1662      * the dynamic sysbus nodes are complex and their generation is not yet
1663      * supported. In that case the user can take charge of the guest dt
1664      * while qemu takes charge of the qom stuff.
1665      */
1666     if (info->dtb_filename == NULL) {
1667         platform_bus_add_all_fdt_nodes(ms->fdt, "/intc",
1668                                        vms->memmap[VIRT_PLATFORM_BUS].base,
1669                                        vms->memmap[VIRT_PLATFORM_BUS].size,
1670                                        vms->irqmap[VIRT_PLATFORM_BUS]);
1671     }
1672     if (arm_load_dtb(info->dtb_start, info, info->dtb_limit, as, ms) < 0) {
1673         exit(1);
1674     }
1675 
1676     fw_cfg_add_extra_pci_roots(vms->bus, vms->fw_cfg);
1677 
1678     virt_acpi_setup(vms);
1679     virt_build_smbios(vms);
1680 }
1681 
1682 static uint64_t virt_cpu_mp_affinity(VirtMachineState *vms, int idx)
1683 {
1684     uint8_t clustersz = ARM_DEFAULT_CPUS_PER_CLUSTER;
1685     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1686 
1687     if (!vmc->disallow_affinity_adjustment) {
1688         /* Adjust MPIDR like 64-bit KVM hosts, which incorporate the
1689          * GIC's target-list limitations. 32-bit KVM hosts currently
1690          * always create clusters of 4 CPUs, but that is expected to
1691          * change when they gain support for gicv3. When KVM is enabled
1692          * it will override the changes we make here, therefore our
1693          * purposes are to make TCG consistent (with 64-bit KVM hosts)
1694          * and to improve SGI efficiency.
1695          */
1696         if (vms->gic_version == VIRT_GIC_VERSION_2) {
1697             clustersz = GIC_TARGETLIST_BITS;
1698         } else {
1699             clustersz = GICV3_TARGETLIST_BITS;
1700         }
1701     }
1702     return arm_cpu_mp_affinity(idx, clustersz);
1703 }
1704 
1705 static inline bool *virt_get_high_memmap_enabled(VirtMachineState *vms,
1706                                                  int index)
1707 {
1708     bool *enabled_array[] = {
1709         &vms->highmem_redists,
1710         &vms->highmem_ecam,
1711         &vms->highmem_mmio,
1712     };
1713 
1714     assert(ARRAY_SIZE(extended_memmap) - VIRT_LOWMEMMAP_LAST ==
1715            ARRAY_SIZE(enabled_array));
1716     assert(index - VIRT_LOWMEMMAP_LAST < ARRAY_SIZE(enabled_array));
1717 
1718     return enabled_array[index - VIRT_LOWMEMMAP_LAST];
1719 }
1720 
1721 static void virt_set_high_memmap(VirtMachineState *vms,
1722                                  hwaddr base, int pa_bits)
1723 {
1724     hwaddr region_base, region_size;
1725     bool *region_enabled, fits;
1726     int i;
1727 
1728     for (i = VIRT_LOWMEMMAP_LAST; i < ARRAY_SIZE(extended_memmap); i++) {
1729         region_enabled = virt_get_high_memmap_enabled(vms, i);
1730         region_base = ROUND_UP(base, extended_memmap[i].size);
1731         region_size = extended_memmap[i].size;
1732 
1733         vms->memmap[i].base = region_base;
1734         vms->memmap[i].size = region_size;
1735 
1736         /*
1737          * Check each device to see if it fits in the PA space,
1738          * moving highest_gpa as we go. For compatibility, move
1739          * highest_gpa for disabled fitting devices as well, if
1740          * the compact layout has been disabled.
1741          *
1742          * For each device that doesn't fit, disable it.
1743          */
1744         fits = (region_base + region_size) <= BIT_ULL(pa_bits);
1745         *region_enabled &= fits;
1746         if (vms->highmem_compact && !*region_enabled) {
1747             continue;
1748         }
1749 
1750         base = region_base + region_size;
1751         if (fits) {
1752             vms->highest_gpa = base - 1;
1753         }
1754     }
1755 }
1756 
1757 static void virt_set_memmap(VirtMachineState *vms, int pa_bits)
1758 {
1759     MachineState *ms = MACHINE(vms);
1760     hwaddr base, device_memory_base, device_memory_size, memtop;
1761     int i;
1762 
1763     vms->memmap = extended_memmap;
1764 
1765     for (i = 0; i < ARRAY_SIZE(base_memmap); i++) {
1766         vms->memmap[i] = base_memmap[i];
1767     }
1768 
1769     if (ms->ram_slots > ACPI_MAX_RAM_SLOTS) {
1770         error_report("unsupported number of memory slots: %"PRIu64,
1771                      ms->ram_slots);
1772         exit(EXIT_FAILURE);
1773     }
1774 
1775     /*
1776      * !highmem is exactly the same as limiting the PA space to 32bit,
1777      * irrespective of the underlying capabilities of the HW.
1778      */
1779     if (!vms->highmem) {
1780         pa_bits = 32;
1781     }
1782 
1783     /*
1784      * We compute the base of the high IO region depending on the
1785      * amount of initial and device memory. The device memory start/size
1786      * is aligned on 1GiB. We never put the high IO region below 256GiB
1787      * so that if maxram_size is < 255GiB we keep the legacy memory map.
1788      * The device region size assumes 1GiB page max alignment per slot.
1789      */
1790     device_memory_base =
1791         ROUND_UP(vms->memmap[VIRT_MEM].base + ms->ram_size, GiB);
1792     device_memory_size = ms->maxram_size - ms->ram_size + ms->ram_slots * GiB;
1793 
1794     /* Base address of the high IO region */
1795     memtop = base = device_memory_base + ROUND_UP(device_memory_size, GiB);
1796     if (memtop > BIT_ULL(pa_bits)) {
1797 	    error_report("Addressing limited to %d bits, but memory exceeds it by %llu bytes\n",
1798 			 pa_bits, memtop - BIT_ULL(pa_bits));
1799         exit(EXIT_FAILURE);
1800     }
1801     if (base < device_memory_base) {
1802         error_report("maxmem/slots too huge");
1803         exit(EXIT_FAILURE);
1804     }
1805     if (base < vms->memmap[VIRT_MEM].base + LEGACY_RAMLIMIT_BYTES) {
1806         base = vms->memmap[VIRT_MEM].base + LEGACY_RAMLIMIT_BYTES;
1807     }
1808 
1809     /* We know for sure that at least the memory fits in the PA space */
1810     vms->highest_gpa = memtop - 1;
1811 
1812     virt_set_high_memmap(vms, base, pa_bits);
1813 
1814     if (device_memory_size > 0) {
1815         machine_memory_devices_init(ms, device_memory_base, device_memory_size);
1816     }
1817 }
1818 
1819 static VirtGICType finalize_gic_version_do(const char *accel_name,
1820                                            VirtGICType gic_version,
1821                                            int gics_supported,
1822                                            unsigned int max_cpus)
1823 {
1824     /* Convert host/max/nosel to GIC version number */
1825     switch (gic_version) {
1826     case VIRT_GIC_VERSION_HOST:
1827         if (!kvm_enabled()) {
1828             error_report("gic-version=host requires KVM");
1829             exit(1);
1830         }
1831 
1832         /* For KVM, gic-version=host means gic-version=max */
1833         return finalize_gic_version_do(accel_name, VIRT_GIC_VERSION_MAX,
1834                                        gics_supported, max_cpus);
1835     case VIRT_GIC_VERSION_MAX:
1836         if (gics_supported & VIRT_GIC_VERSION_4_MASK) {
1837             gic_version = VIRT_GIC_VERSION_4;
1838         } else if (gics_supported & VIRT_GIC_VERSION_3_MASK) {
1839             gic_version = VIRT_GIC_VERSION_3;
1840         } else {
1841             gic_version = VIRT_GIC_VERSION_2;
1842         }
1843         break;
1844     case VIRT_GIC_VERSION_NOSEL:
1845         if ((gics_supported & VIRT_GIC_VERSION_2_MASK) &&
1846             max_cpus <= GIC_NCPU) {
1847             gic_version = VIRT_GIC_VERSION_2;
1848         } else if (gics_supported & VIRT_GIC_VERSION_3_MASK) {
1849             /*
1850              * in case the host does not support v2 emulation or
1851              * the end-user requested more than 8 VCPUs we now default
1852              * to v3. In any case defaulting to v2 would be broken.
1853              */
1854             gic_version = VIRT_GIC_VERSION_3;
1855         } else if (max_cpus > GIC_NCPU) {
1856             error_report("%s only supports GICv2 emulation but more than 8 "
1857                          "vcpus are requested", accel_name);
1858             exit(1);
1859         }
1860         break;
1861     case VIRT_GIC_VERSION_2:
1862     case VIRT_GIC_VERSION_3:
1863     case VIRT_GIC_VERSION_4:
1864         break;
1865     }
1866 
1867     /* Check chosen version is effectively supported */
1868     switch (gic_version) {
1869     case VIRT_GIC_VERSION_2:
1870         if (!(gics_supported & VIRT_GIC_VERSION_2_MASK)) {
1871             error_report("%s does not support GICv2 emulation", accel_name);
1872             exit(1);
1873         }
1874         break;
1875     case VIRT_GIC_VERSION_3:
1876         if (!(gics_supported & VIRT_GIC_VERSION_3_MASK)) {
1877             error_report("%s does not support GICv3 emulation", accel_name);
1878             exit(1);
1879         }
1880         break;
1881     case VIRT_GIC_VERSION_4:
1882         if (!(gics_supported & VIRT_GIC_VERSION_4_MASK)) {
1883             error_report("%s does not support GICv4 emulation, is virtualization=on?",
1884                          accel_name);
1885             exit(1);
1886         }
1887         break;
1888     default:
1889         error_report("logic error in finalize_gic_version");
1890         exit(1);
1891         break;
1892     }
1893 
1894     return gic_version;
1895 }
1896 
1897 /*
1898  * finalize_gic_version - Determines the final gic_version
1899  * according to the gic-version property
1900  *
1901  * Default GIC type is v2
1902  */
1903 static void finalize_gic_version(VirtMachineState *vms)
1904 {
1905     const char *accel_name = current_accel_name();
1906     unsigned int max_cpus = MACHINE(vms)->smp.max_cpus;
1907     int gics_supported = 0;
1908 
1909     /* Determine which GIC versions the current environment supports */
1910     if (kvm_enabled() && kvm_irqchip_in_kernel()) {
1911         int probe_bitmap = kvm_arm_vgic_probe();
1912 
1913         if (!probe_bitmap) {
1914             error_report("Unable to determine GIC version supported by host");
1915             exit(1);
1916         }
1917 
1918         if (probe_bitmap & KVM_ARM_VGIC_V2) {
1919             gics_supported |= VIRT_GIC_VERSION_2_MASK;
1920         }
1921         if (probe_bitmap & KVM_ARM_VGIC_V3) {
1922             gics_supported |= VIRT_GIC_VERSION_3_MASK;
1923         }
1924     } else if (kvm_enabled() && !kvm_irqchip_in_kernel()) {
1925         /* KVM w/o kernel irqchip can only deal with GICv2 */
1926         gics_supported |= VIRT_GIC_VERSION_2_MASK;
1927         accel_name = "KVM with kernel-irqchip=off";
1928     } else if (tcg_enabled() || hvf_enabled() || qtest_enabled())  {
1929         gics_supported |= VIRT_GIC_VERSION_2_MASK;
1930         if (module_object_class_by_name("arm-gicv3")) {
1931             gics_supported |= VIRT_GIC_VERSION_3_MASK;
1932             if (vms->virt) {
1933                 /* GICv4 only makes sense if CPU has EL2 */
1934                 gics_supported |= VIRT_GIC_VERSION_4_MASK;
1935             }
1936         }
1937     } else {
1938         error_report("Unsupported accelerator, can not determine GIC support");
1939         exit(1);
1940     }
1941 
1942     /*
1943      * Then convert helpers like host/max to concrete GIC versions and ensure
1944      * the desired version is supported
1945      */
1946     vms->gic_version = finalize_gic_version_do(accel_name, vms->gic_version,
1947                                                gics_supported, max_cpus);
1948 }
1949 
1950 /*
1951  * virt_cpu_post_init() must be called after the CPUs have
1952  * been realized and the GIC has been created.
1953  */
1954 static void virt_cpu_post_init(VirtMachineState *vms, MemoryRegion *sysmem)
1955 {
1956     int max_cpus = MACHINE(vms)->smp.max_cpus;
1957     bool aarch64, pmu, steal_time;
1958     CPUState *cpu;
1959 
1960     aarch64 = object_property_get_bool(OBJECT(first_cpu), "aarch64", NULL);
1961     pmu = object_property_get_bool(OBJECT(first_cpu), "pmu", NULL);
1962     steal_time = object_property_get_bool(OBJECT(first_cpu),
1963                                           "kvm-steal-time", NULL);
1964 
1965     if (kvm_enabled()) {
1966         hwaddr pvtime_reg_base = vms->memmap[VIRT_PVTIME].base;
1967         hwaddr pvtime_reg_size = vms->memmap[VIRT_PVTIME].size;
1968 
1969         if (steal_time) {
1970             MemoryRegion *pvtime = g_new(MemoryRegion, 1);
1971             hwaddr pvtime_size = max_cpus * PVTIME_SIZE_PER_CPU;
1972 
1973             /* The memory region size must be a multiple of host page size. */
1974             pvtime_size = REAL_HOST_PAGE_ALIGN(pvtime_size);
1975 
1976             if (pvtime_size > pvtime_reg_size) {
1977                 error_report("pvtime requires a %" HWADDR_PRId
1978                              " byte memory region for %d CPUs,"
1979                              " but only %" HWADDR_PRId " has been reserved",
1980                              pvtime_size, max_cpus, pvtime_reg_size);
1981                 exit(1);
1982             }
1983 
1984             memory_region_init_ram(pvtime, NULL, "pvtime", pvtime_size, NULL);
1985             memory_region_add_subregion(sysmem, pvtime_reg_base, pvtime);
1986         }
1987 
1988         CPU_FOREACH(cpu) {
1989             if (pmu) {
1990                 assert(arm_feature(&ARM_CPU(cpu)->env, ARM_FEATURE_PMU));
1991                 if (kvm_irqchip_in_kernel()) {
1992                     kvm_arm_pmu_set_irq(cpu, PPI(VIRTUAL_PMU_IRQ));
1993                 }
1994                 kvm_arm_pmu_init(cpu);
1995             }
1996             if (steal_time) {
1997                 kvm_arm_pvtime_init(cpu, pvtime_reg_base +
1998                                          cpu->cpu_index * PVTIME_SIZE_PER_CPU);
1999             }
2000         }
2001     } else {
2002         if (aarch64 && vms->highmem) {
2003             int requested_pa_size = 64 - clz64(vms->highest_gpa);
2004             int pamax = arm_pamax(ARM_CPU(first_cpu));
2005 
2006             if (pamax < requested_pa_size) {
2007                 error_report("VCPU supports less PA bits (%d) than "
2008                              "requested by the memory map (%d)",
2009                              pamax, requested_pa_size);
2010                 exit(1);
2011             }
2012         }
2013     }
2014 }
2015 
2016 static void machvirt_init(MachineState *machine)
2017 {
2018     VirtMachineState *vms = VIRT_MACHINE(machine);
2019     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(machine);
2020     MachineClass *mc = MACHINE_GET_CLASS(machine);
2021     const CPUArchIdList *possible_cpus;
2022     MemoryRegion *sysmem = get_system_memory();
2023     MemoryRegion *secure_sysmem = NULL;
2024     MemoryRegion *tag_sysmem = NULL;
2025     MemoryRegion *secure_tag_sysmem = NULL;
2026     int n, virt_max_cpus;
2027     bool firmware_loaded;
2028     bool aarch64 = true;
2029     bool has_ged = !vmc->no_ged;
2030     unsigned int smp_cpus = machine->smp.cpus;
2031     unsigned int max_cpus = machine->smp.max_cpus;
2032 
2033     if (!cpu_type_valid(machine->cpu_type)) {
2034         error_report("mach-virt: CPU type %s not supported", machine->cpu_type);
2035         exit(1);
2036     }
2037 
2038     possible_cpus = mc->possible_cpu_arch_ids(machine);
2039 
2040     /*
2041      * In accelerated mode, the memory map is computed earlier in kvm_type()
2042      * to create a VM with the right number of IPA bits.
2043      */
2044     if (!vms->memmap) {
2045         Object *cpuobj;
2046         ARMCPU *armcpu;
2047         int pa_bits;
2048 
2049         /*
2050          * Instantiate a temporary CPU object to find out about what
2051          * we are about to deal with. Once this is done, get rid of
2052          * the object.
2053          */
2054         cpuobj = object_new(possible_cpus->cpus[0].type);
2055         armcpu = ARM_CPU(cpuobj);
2056 
2057         pa_bits = arm_pamax(armcpu);
2058 
2059         object_unref(cpuobj);
2060 
2061         virt_set_memmap(vms, pa_bits);
2062     }
2063 
2064     /* We can probe only here because during property set
2065      * KVM is not available yet
2066      */
2067     finalize_gic_version(vms);
2068 
2069     if (vms->secure) {
2070         /*
2071          * The Secure view of the world is the same as the NonSecure,
2072          * but with a few extra devices. Create it as a container region
2073          * containing the system memory at low priority; any secure-only
2074          * devices go in at higher priority and take precedence.
2075          */
2076         secure_sysmem = g_new(MemoryRegion, 1);
2077         memory_region_init(secure_sysmem, OBJECT(machine), "secure-memory",
2078                            UINT64_MAX);
2079         memory_region_add_subregion_overlap(secure_sysmem, 0, sysmem, -1);
2080     }
2081 
2082     firmware_loaded = virt_firmware_init(vms, sysmem,
2083                                          secure_sysmem ?: sysmem);
2084 
2085     /* If we have an EL3 boot ROM then the assumption is that it will
2086      * implement PSCI itself, so disable QEMU's internal implementation
2087      * so it doesn't get in the way. Instead of starting secondary
2088      * CPUs in PSCI powerdown state we will start them all running and
2089      * let the boot ROM sort them out.
2090      * The usual case is that we do use QEMU's PSCI implementation;
2091      * if the guest has EL2 then we will use SMC as the conduit,
2092      * and otherwise we will use HVC (for backwards compatibility and
2093      * because if we're using KVM then we must use HVC).
2094      */
2095     if (vms->secure && firmware_loaded) {
2096         vms->psci_conduit = QEMU_PSCI_CONDUIT_DISABLED;
2097     } else if (vms->virt) {
2098         vms->psci_conduit = QEMU_PSCI_CONDUIT_SMC;
2099     } else {
2100         vms->psci_conduit = QEMU_PSCI_CONDUIT_HVC;
2101     }
2102 
2103     /*
2104      * The maximum number of CPUs depends on the GIC version, or on how
2105      * many redistributors we can fit into the memory map (which in turn
2106      * depends on whether this is a GICv3 or v4).
2107      */
2108     if (vms->gic_version == VIRT_GIC_VERSION_2) {
2109         virt_max_cpus = GIC_NCPU;
2110     } else {
2111         virt_max_cpus = virt_redist_capacity(vms, VIRT_GIC_REDIST);
2112         if (vms->highmem_redists) {
2113             virt_max_cpus += virt_redist_capacity(vms, VIRT_HIGH_GIC_REDIST2);
2114         }
2115     }
2116 
2117     if (max_cpus > virt_max_cpus) {
2118         error_report("Number of SMP CPUs requested (%d) exceeds max CPUs "
2119                      "supported by machine 'mach-virt' (%d)",
2120                      max_cpus, virt_max_cpus);
2121         if (vms->gic_version != VIRT_GIC_VERSION_2 && !vms->highmem_redists) {
2122             error_printf("Try 'highmem-redists=on' for more CPUs\n");
2123         }
2124 
2125         exit(1);
2126     }
2127 
2128     if (vms->secure && (kvm_enabled() || hvf_enabled())) {
2129         error_report("mach-virt: %s does not support providing "
2130                      "Security extensions (TrustZone) to the guest CPU",
2131                      current_accel_name());
2132         exit(1);
2133     }
2134 
2135     if (vms->virt && (kvm_enabled() || hvf_enabled())) {
2136         error_report("mach-virt: %s does not support providing "
2137                      "Virtualization extensions to the guest CPU",
2138                      current_accel_name());
2139         exit(1);
2140     }
2141 
2142     if (vms->mte && (kvm_enabled() || hvf_enabled())) {
2143         error_report("mach-virt: %s does not support providing "
2144                      "MTE to the guest CPU",
2145                      current_accel_name());
2146         exit(1);
2147     }
2148 
2149     create_fdt(vms);
2150 
2151     assert(possible_cpus->len == max_cpus);
2152     for (n = 0; n < possible_cpus->len; n++) {
2153         Object *cpuobj;
2154         CPUState *cs;
2155 
2156         if (n >= smp_cpus) {
2157             break;
2158         }
2159 
2160         cpuobj = object_new(possible_cpus->cpus[n].type);
2161         object_property_set_int(cpuobj, "mp-affinity",
2162                                 possible_cpus->cpus[n].arch_id, NULL);
2163 
2164         cs = CPU(cpuobj);
2165         cs->cpu_index = n;
2166 
2167         numa_cpu_pre_plug(&possible_cpus->cpus[cs->cpu_index], DEVICE(cpuobj),
2168                           &error_fatal);
2169 
2170         aarch64 &= object_property_get_bool(cpuobj, "aarch64", NULL);
2171 
2172         if (!vms->secure) {
2173             object_property_set_bool(cpuobj, "has_el3", false, NULL);
2174         }
2175 
2176         if (!vms->virt && object_property_find(cpuobj, "has_el2")) {
2177             object_property_set_bool(cpuobj, "has_el2", false, NULL);
2178         }
2179 
2180         if (vmc->kvm_no_adjvtime &&
2181             object_property_find(cpuobj, "kvm-no-adjvtime")) {
2182             object_property_set_bool(cpuobj, "kvm-no-adjvtime", true, NULL);
2183         }
2184 
2185         if (vmc->no_kvm_steal_time &&
2186             object_property_find(cpuobj, "kvm-steal-time")) {
2187             object_property_set_bool(cpuobj, "kvm-steal-time", false, NULL);
2188         }
2189 
2190         if (vmc->no_pmu && object_property_find(cpuobj, "pmu")) {
2191             object_property_set_bool(cpuobj, "pmu", false, NULL);
2192         }
2193 
2194         if (vmc->no_tcg_lpa2 && object_property_find(cpuobj, "lpa2")) {
2195             object_property_set_bool(cpuobj, "lpa2", false, NULL);
2196         }
2197 
2198         if (object_property_find(cpuobj, "reset-cbar")) {
2199             object_property_set_int(cpuobj, "reset-cbar",
2200                                     vms->memmap[VIRT_CPUPERIPHS].base,
2201                                     &error_abort);
2202         }
2203 
2204         object_property_set_link(cpuobj, "memory", OBJECT(sysmem),
2205                                  &error_abort);
2206         if (vms->secure) {
2207             object_property_set_link(cpuobj, "secure-memory",
2208                                      OBJECT(secure_sysmem), &error_abort);
2209         }
2210 
2211         if (vms->mte) {
2212             /* Create the memory region only once, but link to all cpus. */
2213             if (!tag_sysmem) {
2214                 /*
2215                  * The property exists only if MemTag is supported.
2216                  * If it is, we must allocate the ram to back that up.
2217                  */
2218                 if (!object_property_find(cpuobj, "tag-memory")) {
2219                     error_report("MTE requested, but not supported "
2220                                  "by the guest CPU");
2221                     exit(1);
2222                 }
2223 
2224                 tag_sysmem = g_new(MemoryRegion, 1);
2225                 memory_region_init(tag_sysmem, OBJECT(machine),
2226                                    "tag-memory", UINT64_MAX / 32);
2227 
2228                 if (vms->secure) {
2229                     secure_tag_sysmem = g_new(MemoryRegion, 1);
2230                     memory_region_init(secure_tag_sysmem, OBJECT(machine),
2231                                        "secure-tag-memory", UINT64_MAX / 32);
2232 
2233                     /* As with ram, secure-tag takes precedence over tag.  */
2234                     memory_region_add_subregion_overlap(secure_tag_sysmem, 0,
2235                                                         tag_sysmem, -1);
2236                 }
2237             }
2238 
2239             object_property_set_link(cpuobj, "tag-memory", OBJECT(tag_sysmem),
2240                                      &error_abort);
2241             if (vms->secure) {
2242                 object_property_set_link(cpuobj, "secure-tag-memory",
2243                                          OBJECT(secure_tag_sysmem),
2244                                          &error_abort);
2245             }
2246         }
2247 
2248         qdev_realize(DEVICE(cpuobj), NULL, &error_fatal);
2249         object_unref(cpuobj);
2250     }
2251     fdt_add_timer_nodes(vms);
2252     fdt_add_cpu_nodes(vms);
2253 
2254     memory_region_add_subregion(sysmem, vms->memmap[VIRT_MEM].base,
2255                                 machine->ram);
2256 
2257     virt_flash_fdt(vms, sysmem, secure_sysmem ?: sysmem);
2258 
2259     create_gic(vms, sysmem);
2260 
2261     virt_cpu_post_init(vms, sysmem);
2262 
2263     fdt_add_pmu_nodes(vms);
2264 
2265     create_uart(vms, VIRT_UART, sysmem, serial_hd(0));
2266 
2267     if (vms->secure) {
2268         create_secure_ram(vms, secure_sysmem, secure_tag_sysmem);
2269         create_uart(vms, VIRT_SECURE_UART, secure_sysmem, serial_hd(1));
2270     }
2271 
2272     if (tag_sysmem) {
2273         create_tag_ram(tag_sysmem, vms->memmap[VIRT_MEM].base,
2274                        machine->ram_size, "mach-virt.tag");
2275     }
2276 
2277     vms->highmem_ecam &= (!firmware_loaded || aarch64);
2278 
2279     create_rtc(vms);
2280 
2281     create_pcie(vms);
2282 
2283     if (has_ged && aarch64 && firmware_loaded && virt_is_acpi_enabled(vms)) {
2284         vms->acpi_dev = create_acpi_ged(vms);
2285     } else {
2286         create_gpio_devices(vms, VIRT_GPIO, sysmem);
2287     }
2288 
2289     if (vms->secure && !vmc->no_secure_gpio) {
2290         create_gpio_devices(vms, VIRT_SECURE_GPIO, secure_sysmem);
2291     }
2292 
2293      /* connect powerdown request */
2294      vms->powerdown_notifier.notify = virt_powerdown_req;
2295      qemu_register_powerdown_notifier(&vms->powerdown_notifier);
2296 
2297     /* Create mmio transports, so the user can create virtio backends
2298      * (which will be automatically plugged in to the transports). If
2299      * no backend is created the transport will just sit harmlessly idle.
2300      */
2301     create_virtio_devices(vms);
2302 
2303     vms->fw_cfg = create_fw_cfg(vms, &address_space_memory);
2304     rom_set_fw(vms->fw_cfg);
2305 
2306     create_platform_bus(vms);
2307 
2308     if (machine->nvdimms_state->is_enabled) {
2309         const struct AcpiGenericAddress arm_virt_nvdimm_acpi_dsmio = {
2310             .space_id = AML_AS_SYSTEM_MEMORY,
2311             .address = vms->memmap[VIRT_NVDIMM_ACPI].base,
2312             .bit_width = NVDIMM_ACPI_IO_LEN << 3
2313         };
2314 
2315         nvdimm_init_acpi_state(machine->nvdimms_state, sysmem,
2316                                arm_virt_nvdimm_acpi_dsmio,
2317                                vms->fw_cfg, OBJECT(vms));
2318     }
2319 
2320     vms->bootinfo.ram_size = machine->ram_size;
2321     vms->bootinfo.board_id = -1;
2322     vms->bootinfo.loader_start = vms->memmap[VIRT_MEM].base;
2323     vms->bootinfo.get_dtb = machvirt_dtb;
2324     vms->bootinfo.skip_dtb_autoload = true;
2325     vms->bootinfo.firmware_loaded = firmware_loaded;
2326     vms->bootinfo.psci_conduit = vms->psci_conduit;
2327     arm_load_kernel(ARM_CPU(first_cpu), machine, &vms->bootinfo);
2328 
2329     vms->machine_done.notify = virt_machine_done;
2330     qemu_add_machine_init_done_notifier(&vms->machine_done);
2331 }
2332 
2333 static bool virt_get_secure(Object *obj, Error **errp)
2334 {
2335     VirtMachineState *vms = VIRT_MACHINE(obj);
2336 
2337     return vms->secure;
2338 }
2339 
2340 static void virt_set_secure(Object *obj, bool value, Error **errp)
2341 {
2342     VirtMachineState *vms = VIRT_MACHINE(obj);
2343 
2344     vms->secure = value;
2345 }
2346 
2347 static bool virt_get_virt(Object *obj, Error **errp)
2348 {
2349     VirtMachineState *vms = VIRT_MACHINE(obj);
2350 
2351     return vms->virt;
2352 }
2353 
2354 static void virt_set_virt(Object *obj, bool value, Error **errp)
2355 {
2356     VirtMachineState *vms = VIRT_MACHINE(obj);
2357 
2358     vms->virt = value;
2359 }
2360 
2361 static bool virt_get_highmem(Object *obj, Error **errp)
2362 {
2363     VirtMachineState *vms = VIRT_MACHINE(obj);
2364 
2365     return vms->highmem;
2366 }
2367 
2368 static void virt_set_highmem(Object *obj, bool value, Error **errp)
2369 {
2370     VirtMachineState *vms = VIRT_MACHINE(obj);
2371 
2372     vms->highmem = value;
2373 }
2374 
2375 static bool virt_get_compact_highmem(Object *obj, Error **errp)
2376 {
2377     VirtMachineState *vms = VIRT_MACHINE(obj);
2378 
2379     return vms->highmem_compact;
2380 }
2381 
2382 static void virt_set_compact_highmem(Object *obj, bool value, Error **errp)
2383 {
2384     VirtMachineState *vms = VIRT_MACHINE(obj);
2385 
2386     vms->highmem_compact = value;
2387 }
2388 
2389 static bool virt_get_highmem_redists(Object *obj, Error **errp)
2390 {
2391     VirtMachineState *vms = VIRT_MACHINE(obj);
2392 
2393     return vms->highmem_redists;
2394 }
2395 
2396 static void virt_set_highmem_redists(Object *obj, bool value, Error **errp)
2397 {
2398     VirtMachineState *vms = VIRT_MACHINE(obj);
2399 
2400     vms->highmem_redists = value;
2401 }
2402 
2403 static bool virt_get_highmem_ecam(Object *obj, Error **errp)
2404 {
2405     VirtMachineState *vms = VIRT_MACHINE(obj);
2406 
2407     return vms->highmem_ecam;
2408 }
2409 
2410 static void virt_set_highmem_ecam(Object *obj, bool value, Error **errp)
2411 {
2412     VirtMachineState *vms = VIRT_MACHINE(obj);
2413 
2414     vms->highmem_ecam = value;
2415 }
2416 
2417 static bool virt_get_highmem_mmio(Object *obj, Error **errp)
2418 {
2419     VirtMachineState *vms = VIRT_MACHINE(obj);
2420 
2421     return vms->highmem_mmio;
2422 }
2423 
2424 static void virt_set_highmem_mmio(Object *obj, bool value, Error **errp)
2425 {
2426     VirtMachineState *vms = VIRT_MACHINE(obj);
2427 
2428     vms->highmem_mmio = value;
2429 }
2430 
2431 
2432 static bool virt_get_its(Object *obj, Error **errp)
2433 {
2434     VirtMachineState *vms = VIRT_MACHINE(obj);
2435 
2436     return vms->its;
2437 }
2438 
2439 static void virt_set_its(Object *obj, bool value, Error **errp)
2440 {
2441     VirtMachineState *vms = VIRT_MACHINE(obj);
2442 
2443     vms->its = value;
2444 }
2445 
2446 static bool virt_get_dtb_randomness(Object *obj, Error **errp)
2447 {
2448     VirtMachineState *vms = VIRT_MACHINE(obj);
2449 
2450     return vms->dtb_randomness;
2451 }
2452 
2453 static void virt_set_dtb_randomness(Object *obj, bool value, Error **errp)
2454 {
2455     VirtMachineState *vms = VIRT_MACHINE(obj);
2456 
2457     vms->dtb_randomness = value;
2458 }
2459 
2460 static char *virt_get_oem_id(Object *obj, Error **errp)
2461 {
2462     VirtMachineState *vms = VIRT_MACHINE(obj);
2463 
2464     return g_strdup(vms->oem_id);
2465 }
2466 
2467 static void virt_set_oem_id(Object *obj, const char *value, Error **errp)
2468 {
2469     VirtMachineState *vms = VIRT_MACHINE(obj);
2470     size_t len = strlen(value);
2471 
2472     if (len > 6) {
2473         error_setg(errp,
2474                    "User specified oem-id value is bigger than 6 bytes in size");
2475         return;
2476     }
2477 
2478     strncpy(vms->oem_id, value, 6);
2479 }
2480 
2481 static char *virt_get_oem_table_id(Object *obj, Error **errp)
2482 {
2483     VirtMachineState *vms = VIRT_MACHINE(obj);
2484 
2485     return g_strdup(vms->oem_table_id);
2486 }
2487 
2488 static void virt_set_oem_table_id(Object *obj, const char *value,
2489                                   Error **errp)
2490 {
2491     VirtMachineState *vms = VIRT_MACHINE(obj);
2492     size_t len = strlen(value);
2493 
2494     if (len > 8) {
2495         error_setg(errp,
2496                    "User specified oem-table-id value is bigger than 8 bytes in size");
2497         return;
2498     }
2499     strncpy(vms->oem_table_id, value, 8);
2500 }
2501 
2502 
2503 bool virt_is_acpi_enabled(VirtMachineState *vms)
2504 {
2505     if (vms->acpi == ON_OFF_AUTO_OFF) {
2506         return false;
2507     }
2508     return true;
2509 }
2510 
2511 static void virt_get_acpi(Object *obj, Visitor *v, const char *name,
2512                           void *opaque, Error **errp)
2513 {
2514     VirtMachineState *vms = VIRT_MACHINE(obj);
2515     OnOffAuto acpi = vms->acpi;
2516 
2517     visit_type_OnOffAuto(v, name, &acpi, errp);
2518 }
2519 
2520 static void virt_set_acpi(Object *obj, Visitor *v, const char *name,
2521                           void *opaque, Error **errp)
2522 {
2523     VirtMachineState *vms = VIRT_MACHINE(obj);
2524 
2525     visit_type_OnOffAuto(v, name, &vms->acpi, errp);
2526 }
2527 
2528 static bool virt_get_ras(Object *obj, Error **errp)
2529 {
2530     VirtMachineState *vms = VIRT_MACHINE(obj);
2531 
2532     return vms->ras;
2533 }
2534 
2535 static void virt_set_ras(Object *obj, bool value, Error **errp)
2536 {
2537     VirtMachineState *vms = VIRT_MACHINE(obj);
2538 
2539     vms->ras = value;
2540 }
2541 
2542 static bool virt_get_mte(Object *obj, Error **errp)
2543 {
2544     VirtMachineState *vms = VIRT_MACHINE(obj);
2545 
2546     return vms->mte;
2547 }
2548 
2549 static void virt_set_mte(Object *obj, bool value, Error **errp)
2550 {
2551     VirtMachineState *vms = VIRT_MACHINE(obj);
2552 
2553     vms->mte = value;
2554 }
2555 
2556 static char *virt_get_gic_version(Object *obj, Error **errp)
2557 {
2558     VirtMachineState *vms = VIRT_MACHINE(obj);
2559     const char *val;
2560 
2561     switch (vms->gic_version) {
2562     case VIRT_GIC_VERSION_4:
2563         val = "4";
2564         break;
2565     case VIRT_GIC_VERSION_3:
2566         val = "3";
2567         break;
2568     default:
2569         val = "2";
2570         break;
2571     }
2572     return g_strdup(val);
2573 }
2574 
2575 static void virt_set_gic_version(Object *obj, const char *value, Error **errp)
2576 {
2577     VirtMachineState *vms = VIRT_MACHINE(obj);
2578 
2579     if (!strcmp(value, "4")) {
2580         vms->gic_version = VIRT_GIC_VERSION_4;
2581     } else if (!strcmp(value, "3")) {
2582         vms->gic_version = VIRT_GIC_VERSION_3;
2583     } else if (!strcmp(value, "2")) {
2584         vms->gic_version = VIRT_GIC_VERSION_2;
2585     } else if (!strcmp(value, "host")) {
2586         vms->gic_version = VIRT_GIC_VERSION_HOST; /* Will probe later */
2587     } else if (!strcmp(value, "max")) {
2588         vms->gic_version = VIRT_GIC_VERSION_MAX; /* Will probe later */
2589     } else {
2590         error_setg(errp, "Invalid gic-version value");
2591         error_append_hint(errp, "Valid values are 3, 2, host, max.\n");
2592     }
2593 }
2594 
2595 static char *virt_get_iommu(Object *obj, Error **errp)
2596 {
2597     VirtMachineState *vms = VIRT_MACHINE(obj);
2598 
2599     switch (vms->iommu) {
2600     case VIRT_IOMMU_NONE:
2601         return g_strdup("none");
2602     case VIRT_IOMMU_SMMUV3:
2603         return g_strdup("smmuv3");
2604     default:
2605         g_assert_not_reached();
2606     }
2607 }
2608 
2609 static void virt_set_iommu(Object *obj, const char *value, Error **errp)
2610 {
2611     VirtMachineState *vms = VIRT_MACHINE(obj);
2612 
2613     if (!strcmp(value, "smmuv3")) {
2614         vms->iommu = VIRT_IOMMU_SMMUV3;
2615     } else if (!strcmp(value, "none")) {
2616         vms->iommu = VIRT_IOMMU_NONE;
2617     } else {
2618         error_setg(errp, "Invalid iommu value");
2619         error_append_hint(errp, "Valid values are none, smmuv3.\n");
2620     }
2621 }
2622 
2623 static bool virt_get_default_bus_bypass_iommu(Object *obj, Error **errp)
2624 {
2625     VirtMachineState *vms = VIRT_MACHINE(obj);
2626 
2627     return vms->default_bus_bypass_iommu;
2628 }
2629 
2630 static void virt_set_default_bus_bypass_iommu(Object *obj, bool value,
2631                                               Error **errp)
2632 {
2633     VirtMachineState *vms = VIRT_MACHINE(obj);
2634 
2635     vms->default_bus_bypass_iommu = value;
2636 }
2637 
2638 static CpuInstanceProperties
2639 virt_cpu_index_to_props(MachineState *ms, unsigned cpu_index)
2640 {
2641     MachineClass *mc = MACHINE_GET_CLASS(ms);
2642     const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms);
2643 
2644     assert(cpu_index < possible_cpus->len);
2645     return possible_cpus->cpus[cpu_index].props;
2646 }
2647 
2648 static int64_t virt_get_default_cpu_node_id(const MachineState *ms, int idx)
2649 {
2650     int64_t socket_id = ms->possible_cpus->cpus[idx].props.socket_id;
2651 
2652     return socket_id % ms->numa_state->num_nodes;
2653 }
2654 
2655 static const CPUArchIdList *virt_possible_cpu_arch_ids(MachineState *ms)
2656 {
2657     int n;
2658     unsigned int max_cpus = ms->smp.max_cpus;
2659     VirtMachineState *vms = VIRT_MACHINE(ms);
2660     MachineClass *mc = MACHINE_GET_CLASS(vms);
2661 
2662     if (ms->possible_cpus) {
2663         assert(ms->possible_cpus->len == max_cpus);
2664         return ms->possible_cpus;
2665     }
2666 
2667     ms->possible_cpus = g_malloc0(sizeof(CPUArchIdList) +
2668                                   sizeof(CPUArchId) * max_cpus);
2669     ms->possible_cpus->len = max_cpus;
2670     for (n = 0; n < ms->possible_cpus->len; n++) {
2671         ms->possible_cpus->cpus[n].type = ms->cpu_type;
2672         ms->possible_cpus->cpus[n].arch_id =
2673             virt_cpu_mp_affinity(vms, n);
2674 
2675         assert(!mc->smp_props.dies_supported);
2676         ms->possible_cpus->cpus[n].props.has_socket_id = true;
2677         ms->possible_cpus->cpus[n].props.socket_id =
2678             n / (ms->smp.clusters * ms->smp.cores * ms->smp.threads);
2679         ms->possible_cpus->cpus[n].props.has_cluster_id = true;
2680         ms->possible_cpus->cpus[n].props.cluster_id =
2681             (n / (ms->smp.cores * ms->smp.threads)) % ms->smp.clusters;
2682         ms->possible_cpus->cpus[n].props.has_core_id = true;
2683         ms->possible_cpus->cpus[n].props.core_id =
2684             (n / ms->smp.threads) % ms->smp.cores;
2685         ms->possible_cpus->cpus[n].props.has_thread_id = true;
2686         ms->possible_cpus->cpus[n].props.thread_id =
2687             n % ms->smp.threads;
2688     }
2689     return ms->possible_cpus;
2690 }
2691 
2692 static void virt_memory_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
2693                                  Error **errp)
2694 {
2695     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2696     const MachineState *ms = MACHINE(hotplug_dev);
2697     const bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2698 
2699     if (!vms->acpi_dev) {
2700         error_setg(errp,
2701                    "memory hotplug is not enabled: missing acpi-ged device");
2702         return;
2703     }
2704 
2705     if (vms->mte) {
2706         error_setg(errp, "memory hotplug is not enabled: MTE is enabled");
2707         return;
2708     }
2709 
2710     if (is_nvdimm && !ms->nvdimms_state->is_enabled) {
2711         error_setg(errp, "nvdimm is not enabled: add 'nvdimm=on' to '-M'");
2712         return;
2713     }
2714 
2715     pc_dimm_pre_plug(PC_DIMM(dev), MACHINE(hotplug_dev), NULL, errp);
2716 }
2717 
2718 static void virt_memory_plug(HotplugHandler *hotplug_dev,
2719                              DeviceState *dev, Error **errp)
2720 {
2721     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2722     MachineState *ms = MACHINE(hotplug_dev);
2723     bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2724 
2725     pc_dimm_plug(PC_DIMM(dev), MACHINE(vms));
2726 
2727     if (is_nvdimm) {
2728         nvdimm_plug(ms->nvdimms_state);
2729     }
2730 
2731     hotplug_handler_plug(HOTPLUG_HANDLER(vms->acpi_dev),
2732                          dev, &error_abort);
2733 }
2734 
2735 static void virt_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev,
2736                                             DeviceState *dev, Error **errp)
2737 {
2738     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2739 
2740     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2741         virt_memory_pre_plug(hotplug_dev, dev, errp);
2742     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI)) {
2743         virtio_md_pci_pre_plug(VIRTIO_MD_PCI(dev), MACHINE(hotplug_dev), errp);
2744     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2745         hwaddr db_start = 0, db_end = 0;
2746         char *resv_prop_str;
2747 
2748         if (vms->iommu != VIRT_IOMMU_NONE) {
2749             error_setg(errp, "virt machine does not support multiple IOMMUs");
2750             return;
2751         }
2752 
2753         switch (vms->msi_controller) {
2754         case VIRT_MSI_CTRL_NONE:
2755             return;
2756         case VIRT_MSI_CTRL_ITS:
2757             /* GITS_TRANSLATER page */
2758             db_start = base_memmap[VIRT_GIC_ITS].base + 0x10000;
2759             db_end = base_memmap[VIRT_GIC_ITS].base +
2760                      base_memmap[VIRT_GIC_ITS].size - 1;
2761             break;
2762         case VIRT_MSI_CTRL_GICV2M:
2763             /* MSI_SETSPI_NS page */
2764             db_start = base_memmap[VIRT_GIC_V2M].base;
2765             db_end = db_start + base_memmap[VIRT_GIC_V2M].size - 1;
2766             break;
2767         }
2768         resv_prop_str = g_strdup_printf("0x%"PRIx64":0x%"PRIx64":%u",
2769                                         db_start, db_end,
2770                                         VIRTIO_IOMMU_RESV_MEM_T_MSI);
2771 
2772         object_property_set_uint(OBJECT(dev), "len-reserved-regions", 1, errp);
2773         object_property_set_str(OBJECT(dev), "reserved-regions[0]",
2774                                 resv_prop_str, errp);
2775         g_free(resv_prop_str);
2776     }
2777 }
2778 
2779 static void virt_machine_device_plug_cb(HotplugHandler *hotplug_dev,
2780                                         DeviceState *dev, Error **errp)
2781 {
2782     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2783 
2784     if (vms->platform_bus_dev) {
2785         MachineClass *mc = MACHINE_GET_CLASS(vms);
2786 
2787         if (device_is_dynamic_sysbus(mc, dev)) {
2788             platform_bus_link_device(PLATFORM_BUS_DEVICE(vms->platform_bus_dev),
2789                                      SYS_BUS_DEVICE(dev));
2790         }
2791     }
2792 
2793     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2794         virt_memory_plug(hotplug_dev, dev, errp);
2795     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI)) {
2796         virtio_md_pci_plug(VIRTIO_MD_PCI(dev), MACHINE(hotplug_dev), errp);
2797     }
2798 
2799     if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2800         PCIDevice *pdev = PCI_DEVICE(dev);
2801 
2802         vms->iommu = VIRT_IOMMU_VIRTIO;
2803         vms->virtio_iommu_bdf = pci_get_bdf(pdev);
2804         create_virtio_iommu_dt_bindings(vms);
2805     }
2806 }
2807 
2808 static void virt_dimm_unplug_request(HotplugHandler *hotplug_dev,
2809                                      DeviceState *dev, Error **errp)
2810 {
2811     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2812 
2813     if (!vms->acpi_dev) {
2814         error_setg(errp,
2815                    "memory hotplug is not enabled: missing acpi-ged device");
2816         return;
2817     }
2818 
2819     if (object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM)) {
2820         error_setg(errp, "nvdimm device hot unplug is not supported yet.");
2821         return;
2822     }
2823 
2824     hotplug_handler_unplug_request(HOTPLUG_HANDLER(vms->acpi_dev), dev,
2825                                    errp);
2826 }
2827 
2828 static void virt_dimm_unplug(HotplugHandler *hotplug_dev,
2829                              DeviceState *dev, Error **errp)
2830 {
2831     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2832     Error *local_err = NULL;
2833 
2834     hotplug_handler_unplug(HOTPLUG_HANDLER(vms->acpi_dev), dev, &local_err);
2835     if (local_err) {
2836         goto out;
2837     }
2838 
2839     pc_dimm_unplug(PC_DIMM(dev), MACHINE(vms));
2840     qdev_unrealize(dev);
2841 
2842 out:
2843     error_propagate(errp, local_err);
2844 }
2845 
2846 static void virt_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
2847                                           DeviceState *dev, Error **errp)
2848 {
2849     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2850         virt_dimm_unplug_request(hotplug_dev, dev, errp);
2851     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI)) {
2852         virtio_md_pci_unplug_request(VIRTIO_MD_PCI(dev), MACHINE(hotplug_dev),
2853                                      errp);
2854     } else {
2855         error_setg(errp, "device unplug request for unsupported device"
2856                    " type: %s", object_get_typename(OBJECT(dev)));
2857     }
2858 }
2859 
2860 static void virt_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
2861                                           DeviceState *dev, Error **errp)
2862 {
2863     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2864         virt_dimm_unplug(hotplug_dev, dev, errp);
2865     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI)) {
2866         virtio_md_pci_unplug(VIRTIO_MD_PCI(dev), MACHINE(hotplug_dev), errp);
2867     } else {
2868         error_setg(errp, "virt: device unplug for unsupported device"
2869                    " type: %s", object_get_typename(OBJECT(dev)));
2870     }
2871 }
2872 
2873 static HotplugHandler *virt_machine_get_hotplug_handler(MachineState *machine,
2874                                                         DeviceState *dev)
2875 {
2876     MachineClass *mc = MACHINE_GET_CLASS(machine);
2877 
2878     if (device_is_dynamic_sysbus(mc, dev) ||
2879         object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) ||
2880         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI) ||
2881         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2882         return HOTPLUG_HANDLER(machine);
2883     }
2884     return NULL;
2885 }
2886 
2887 /*
2888  * for arm64 kvm_type [7-0] encodes the requested number of bits
2889  * in the IPA address space
2890  */
2891 static int virt_kvm_type(MachineState *ms, const char *type_str)
2892 {
2893     VirtMachineState *vms = VIRT_MACHINE(ms);
2894     int max_vm_pa_size, requested_pa_size;
2895     bool fixed_ipa;
2896 
2897     max_vm_pa_size = kvm_arm_get_max_vm_ipa_size(ms, &fixed_ipa);
2898 
2899     /* we freeze the memory map to compute the highest gpa */
2900     virt_set_memmap(vms, max_vm_pa_size);
2901 
2902     requested_pa_size = 64 - clz64(vms->highest_gpa);
2903 
2904     /*
2905      * KVM requires the IPA size to be at least 32 bits.
2906      */
2907     if (requested_pa_size < 32) {
2908         requested_pa_size = 32;
2909     }
2910 
2911     if (requested_pa_size > max_vm_pa_size) {
2912         error_report("-m and ,maxmem option values "
2913                      "require an IPA range (%d bits) larger than "
2914                      "the one supported by the host (%d bits)",
2915                      requested_pa_size, max_vm_pa_size);
2916         exit(1);
2917     }
2918     /*
2919      * We return the requested PA log size, unless KVM only supports
2920      * the implicit legacy 40b IPA setting, in which case the kvm_type
2921      * must be 0.
2922      */
2923     return fixed_ipa ? 0 : requested_pa_size;
2924 }
2925 
2926 static void virt_machine_class_init(ObjectClass *oc, void *data)
2927 {
2928     MachineClass *mc = MACHINE_CLASS(oc);
2929     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
2930 
2931     mc->init = machvirt_init;
2932     /* Start with max_cpus set to 512, which is the maximum supported by KVM.
2933      * The value may be reduced later when we have more information about the
2934      * configuration of the particular instance.
2935      */
2936     mc->max_cpus = 512;
2937     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_CALXEDA_XGMAC);
2938     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_AMD_XGBE);
2939     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_RAMFB_DEVICE);
2940     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_PLATFORM);
2941 #ifdef CONFIG_TPM
2942     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_TPM_TIS_SYSBUS);
2943 #endif
2944     mc->block_default_type = IF_VIRTIO;
2945     mc->no_cdrom = 1;
2946     mc->pci_allow_0_address = true;
2947     /* We know we will never create a pre-ARMv7 CPU which needs 1K pages */
2948     mc->minimum_page_bits = 12;
2949     mc->possible_cpu_arch_ids = virt_possible_cpu_arch_ids;
2950     mc->cpu_index_to_instance_props = virt_cpu_index_to_props;
2951 #ifdef CONFIG_TCG
2952     mc->default_cpu_type = ARM_CPU_TYPE_NAME("cortex-a15");
2953 #else
2954     mc->default_cpu_type = ARM_CPU_TYPE_NAME("max");
2955 #endif
2956     mc->get_default_cpu_node_id = virt_get_default_cpu_node_id;
2957     mc->kvm_type = virt_kvm_type;
2958     assert(!mc->get_hotplug_handler);
2959     mc->get_hotplug_handler = virt_machine_get_hotplug_handler;
2960     hc->pre_plug = virt_machine_device_pre_plug_cb;
2961     hc->plug = virt_machine_device_plug_cb;
2962     hc->unplug_request = virt_machine_device_unplug_request_cb;
2963     hc->unplug = virt_machine_device_unplug_cb;
2964     mc->nvdimm_supported = true;
2965     mc->smp_props.clusters_supported = true;
2966     mc->auto_enable_numa_with_memhp = true;
2967     mc->auto_enable_numa_with_memdev = true;
2968     /* platform instead of architectural choice */
2969     mc->cpu_cluster_has_numa_boundary = true;
2970     mc->default_ram_id = "mach-virt.ram";
2971     mc->default_nic = "virtio-net-pci";
2972 
2973     object_class_property_add(oc, "acpi", "OnOffAuto",
2974         virt_get_acpi, virt_set_acpi,
2975         NULL, NULL);
2976     object_class_property_set_description(oc, "acpi",
2977         "Enable ACPI");
2978     object_class_property_add_bool(oc, "secure", virt_get_secure,
2979                                    virt_set_secure);
2980     object_class_property_set_description(oc, "secure",
2981                                                 "Set on/off to enable/disable the ARM "
2982                                                 "Security Extensions (TrustZone)");
2983 
2984     object_class_property_add_bool(oc, "virtualization", virt_get_virt,
2985                                    virt_set_virt);
2986     object_class_property_set_description(oc, "virtualization",
2987                                           "Set on/off to enable/disable emulating a "
2988                                           "guest CPU which implements the ARM "
2989                                           "Virtualization Extensions");
2990 
2991     object_class_property_add_bool(oc, "highmem", virt_get_highmem,
2992                                    virt_set_highmem);
2993     object_class_property_set_description(oc, "highmem",
2994                                           "Set on/off to enable/disable using "
2995                                           "physical address space above 32 bits");
2996 
2997     object_class_property_add_bool(oc, "compact-highmem",
2998                                    virt_get_compact_highmem,
2999                                    virt_set_compact_highmem);
3000     object_class_property_set_description(oc, "compact-highmem",
3001                                           "Set on/off to enable/disable compact "
3002                                           "layout for high memory regions");
3003 
3004     object_class_property_add_bool(oc, "highmem-redists",
3005                                    virt_get_highmem_redists,
3006                                    virt_set_highmem_redists);
3007     object_class_property_set_description(oc, "highmem-redists",
3008                                           "Set on/off to enable/disable high "
3009                                           "memory region for GICv3 or GICv4 "
3010                                           "redistributor");
3011 
3012     object_class_property_add_bool(oc, "highmem-ecam",
3013                                    virt_get_highmem_ecam,
3014                                    virt_set_highmem_ecam);
3015     object_class_property_set_description(oc, "highmem-ecam",
3016                                           "Set on/off to enable/disable high "
3017                                           "memory region for PCI ECAM");
3018 
3019     object_class_property_add_bool(oc, "highmem-mmio",
3020                                    virt_get_highmem_mmio,
3021                                    virt_set_highmem_mmio);
3022     object_class_property_set_description(oc, "highmem-mmio",
3023                                           "Set on/off to enable/disable high "
3024                                           "memory region for PCI MMIO");
3025 
3026     object_class_property_add_str(oc, "gic-version", virt_get_gic_version,
3027                                   virt_set_gic_version);
3028     object_class_property_set_description(oc, "gic-version",
3029                                           "Set GIC version. "
3030                                           "Valid values are 2, 3, 4, host and max");
3031 
3032     object_class_property_add_str(oc, "iommu", virt_get_iommu, virt_set_iommu);
3033     object_class_property_set_description(oc, "iommu",
3034                                           "Set the IOMMU type. "
3035                                           "Valid values are none and smmuv3");
3036 
3037     object_class_property_add_bool(oc, "default-bus-bypass-iommu",
3038                                    virt_get_default_bus_bypass_iommu,
3039                                    virt_set_default_bus_bypass_iommu);
3040     object_class_property_set_description(oc, "default-bus-bypass-iommu",
3041                                           "Set on/off to enable/disable "
3042                                           "bypass_iommu for default root bus");
3043 
3044     object_class_property_add_bool(oc, "ras", virt_get_ras,
3045                                    virt_set_ras);
3046     object_class_property_set_description(oc, "ras",
3047                                           "Set on/off to enable/disable reporting host memory errors "
3048                                           "to a KVM guest using ACPI and guest external abort exceptions");
3049 
3050     object_class_property_add_bool(oc, "mte", virt_get_mte, virt_set_mte);
3051     object_class_property_set_description(oc, "mte",
3052                                           "Set on/off to enable/disable emulating a "
3053                                           "guest CPU which implements the ARM "
3054                                           "Memory Tagging Extension");
3055 
3056     object_class_property_add_bool(oc, "its", virt_get_its,
3057                                    virt_set_its);
3058     object_class_property_set_description(oc, "its",
3059                                           "Set on/off to enable/disable "
3060                                           "ITS instantiation");
3061 
3062     object_class_property_add_bool(oc, "dtb-randomness",
3063                                    virt_get_dtb_randomness,
3064                                    virt_set_dtb_randomness);
3065     object_class_property_set_description(oc, "dtb-randomness",
3066                                           "Set off to disable passing random or "
3067                                           "non-deterministic dtb nodes to guest");
3068 
3069     object_class_property_add_bool(oc, "dtb-kaslr-seed",
3070                                    virt_get_dtb_randomness,
3071                                    virt_set_dtb_randomness);
3072     object_class_property_set_description(oc, "dtb-kaslr-seed",
3073                                           "Deprecated synonym of dtb-randomness");
3074 
3075     object_class_property_add_str(oc, "x-oem-id",
3076                                   virt_get_oem_id,
3077                                   virt_set_oem_id);
3078     object_class_property_set_description(oc, "x-oem-id",
3079                                           "Override the default value of field OEMID "
3080                                           "in ACPI table header."
3081                                           "The string may be up to 6 bytes in size");
3082 
3083 
3084     object_class_property_add_str(oc, "x-oem-table-id",
3085                                   virt_get_oem_table_id,
3086                                   virt_set_oem_table_id);
3087     object_class_property_set_description(oc, "x-oem-table-id",
3088                                           "Override the default value of field OEM Table ID "
3089                                           "in ACPI table header."
3090                                           "The string may be up to 8 bytes in size");
3091 
3092 }
3093 
3094 static void virt_instance_init(Object *obj)
3095 {
3096     VirtMachineState *vms = VIRT_MACHINE(obj);
3097     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
3098 
3099     /* EL3 is disabled by default on virt: this makes us consistent
3100      * between KVM and TCG for this board, and it also allows us to
3101      * boot UEFI blobs which assume no TrustZone support.
3102      */
3103     vms->secure = false;
3104 
3105     /* EL2 is also disabled by default, for similar reasons */
3106     vms->virt = false;
3107 
3108     /* High memory is enabled by default */
3109     vms->highmem = true;
3110     vms->highmem_compact = !vmc->no_highmem_compact;
3111     vms->gic_version = VIRT_GIC_VERSION_NOSEL;
3112 
3113     vms->highmem_ecam = !vmc->no_highmem_ecam;
3114     vms->highmem_mmio = true;
3115     vms->highmem_redists = true;
3116 
3117     if (vmc->no_its) {
3118         vms->its = false;
3119     } else {
3120         /* Default allows ITS instantiation */
3121         vms->its = true;
3122 
3123         if (vmc->no_tcg_its) {
3124             vms->tcg_its = false;
3125         } else {
3126             vms->tcg_its = true;
3127         }
3128     }
3129 
3130     /* Default disallows iommu instantiation */
3131     vms->iommu = VIRT_IOMMU_NONE;
3132 
3133     /* The default root bus is attached to iommu by default */
3134     vms->default_bus_bypass_iommu = false;
3135 
3136     /* Default disallows RAS instantiation */
3137     vms->ras = false;
3138 
3139     /* MTE is disabled by default.  */
3140     vms->mte = false;
3141 
3142     /* Supply kaslr-seed and rng-seed by default */
3143     vms->dtb_randomness = true;
3144 
3145     vms->irqmap = a15irqmap;
3146 
3147     virt_flash_create(vms);
3148 
3149     vms->oem_id = g_strndup(ACPI_BUILD_APPNAME6, 6);
3150     vms->oem_table_id = g_strndup(ACPI_BUILD_APPNAME8, 8);
3151 }
3152 
3153 static const TypeInfo virt_machine_info = {
3154     .name          = TYPE_VIRT_MACHINE,
3155     .parent        = TYPE_MACHINE,
3156     .abstract      = true,
3157     .instance_size = sizeof(VirtMachineState),
3158     .class_size    = sizeof(VirtMachineClass),
3159     .class_init    = virt_machine_class_init,
3160     .instance_init = virt_instance_init,
3161     .interfaces = (InterfaceInfo[]) {
3162          { TYPE_HOTPLUG_HANDLER },
3163          { }
3164     },
3165 };
3166 
3167 static void machvirt_machine_init(void)
3168 {
3169     type_register_static(&virt_machine_info);
3170 }
3171 type_init(machvirt_machine_init);
3172 
3173 static void virt_machine_8_1_options(MachineClass *mc)
3174 {
3175 }
3176 DEFINE_VIRT_MACHINE_AS_LATEST(8, 1)
3177 
3178 static void virt_machine_8_0_options(MachineClass *mc)
3179 {
3180     virt_machine_8_1_options(mc);
3181     compat_props_add(mc->compat_props, hw_compat_8_0, hw_compat_8_0_len);
3182 }
3183 DEFINE_VIRT_MACHINE(8, 0)
3184 
3185 static void virt_machine_7_2_options(MachineClass *mc)
3186 {
3187     virt_machine_8_0_options(mc);
3188     compat_props_add(mc->compat_props, hw_compat_7_2, hw_compat_7_2_len);
3189 }
3190 DEFINE_VIRT_MACHINE(7, 2)
3191 
3192 static void virt_machine_7_1_options(MachineClass *mc)
3193 {
3194     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3195 
3196     virt_machine_7_2_options(mc);
3197     compat_props_add(mc->compat_props, hw_compat_7_1, hw_compat_7_1_len);
3198     /* Compact layout for high memory regions was introduced with 7.2 */
3199     vmc->no_highmem_compact = true;
3200 }
3201 DEFINE_VIRT_MACHINE(7, 1)
3202 
3203 static void virt_machine_7_0_options(MachineClass *mc)
3204 {
3205     virt_machine_7_1_options(mc);
3206     compat_props_add(mc->compat_props, hw_compat_7_0, hw_compat_7_0_len);
3207 }
3208 DEFINE_VIRT_MACHINE(7, 0)
3209 
3210 static void virt_machine_6_2_options(MachineClass *mc)
3211 {
3212     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3213 
3214     virt_machine_7_0_options(mc);
3215     compat_props_add(mc->compat_props, hw_compat_6_2, hw_compat_6_2_len);
3216     vmc->no_tcg_lpa2 = true;
3217 }
3218 DEFINE_VIRT_MACHINE(6, 2)
3219 
3220 static void virt_machine_6_1_options(MachineClass *mc)
3221 {
3222     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3223 
3224     virt_machine_6_2_options(mc);
3225     compat_props_add(mc->compat_props, hw_compat_6_1, hw_compat_6_1_len);
3226     mc->smp_props.prefer_sockets = true;
3227     vmc->no_cpu_topology = true;
3228 
3229     /* qemu ITS was introduced with 6.2 */
3230     vmc->no_tcg_its = true;
3231 }
3232 DEFINE_VIRT_MACHINE(6, 1)
3233 
3234 static void virt_machine_6_0_options(MachineClass *mc)
3235 {
3236     virt_machine_6_1_options(mc);
3237     compat_props_add(mc->compat_props, hw_compat_6_0, hw_compat_6_0_len);
3238 }
3239 DEFINE_VIRT_MACHINE(6, 0)
3240 
3241 static void virt_machine_5_2_options(MachineClass *mc)
3242 {
3243     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3244 
3245     virt_machine_6_0_options(mc);
3246     compat_props_add(mc->compat_props, hw_compat_5_2, hw_compat_5_2_len);
3247     vmc->no_secure_gpio = true;
3248 }
3249 DEFINE_VIRT_MACHINE(5, 2)
3250 
3251 static void virt_machine_5_1_options(MachineClass *mc)
3252 {
3253     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3254 
3255     virt_machine_5_2_options(mc);
3256     compat_props_add(mc->compat_props, hw_compat_5_1, hw_compat_5_1_len);
3257     vmc->no_kvm_steal_time = true;
3258 }
3259 DEFINE_VIRT_MACHINE(5, 1)
3260 
3261 static void virt_machine_5_0_options(MachineClass *mc)
3262 {
3263     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3264 
3265     virt_machine_5_1_options(mc);
3266     compat_props_add(mc->compat_props, hw_compat_5_0, hw_compat_5_0_len);
3267     mc->numa_mem_supported = true;
3268     vmc->acpi_expose_flash = true;
3269     mc->auto_enable_numa_with_memdev = false;
3270 }
3271 DEFINE_VIRT_MACHINE(5, 0)
3272 
3273 static void virt_machine_4_2_options(MachineClass *mc)
3274 {
3275     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3276 
3277     virt_machine_5_0_options(mc);
3278     compat_props_add(mc->compat_props, hw_compat_4_2, hw_compat_4_2_len);
3279     vmc->kvm_no_adjvtime = true;
3280 }
3281 DEFINE_VIRT_MACHINE(4, 2)
3282 
3283 static void virt_machine_4_1_options(MachineClass *mc)
3284 {
3285     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3286 
3287     virt_machine_4_2_options(mc);
3288     compat_props_add(mc->compat_props, hw_compat_4_1, hw_compat_4_1_len);
3289     vmc->no_ged = true;
3290     mc->auto_enable_numa_with_memhp = false;
3291 }
3292 DEFINE_VIRT_MACHINE(4, 1)
3293 
3294 static void virt_machine_4_0_options(MachineClass *mc)
3295 {
3296     virt_machine_4_1_options(mc);
3297     compat_props_add(mc->compat_props, hw_compat_4_0, hw_compat_4_0_len);
3298 }
3299 DEFINE_VIRT_MACHINE(4, 0)
3300 
3301 static void virt_machine_3_1_options(MachineClass *mc)
3302 {
3303     virt_machine_4_0_options(mc);
3304     compat_props_add(mc->compat_props, hw_compat_3_1, hw_compat_3_1_len);
3305 }
3306 DEFINE_VIRT_MACHINE(3, 1)
3307 
3308 static void virt_machine_3_0_options(MachineClass *mc)
3309 {
3310     virt_machine_3_1_options(mc);
3311     compat_props_add(mc->compat_props, hw_compat_3_0, hw_compat_3_0_len);
3312 }
3313 DEFINE_VIRT_MACHINE(3, 0)
3314 
3315 static void virt_machine_2_12_options(MachineClass *mc)
3316 {
3317     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3318 
3319     virt_machine_3_0_options(mc);
3320     compat_props_add(mc->compat_props, hw_compat_2_12, hw_compat_2_12_len);
3321     vmc->no_highmem_ecam = true;
3322     mc->max_cpus = 255;
3323 }
3324 DEFINE_VIRT_MACHINE(2, 12)
3325 
3326 static void virt_machine_2_11_options(MachineClass *mc)
3327 {
3328     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3329 
3330     virt_machine_2_12_options(mc);
3331     compat_props_add(mc->compat_props, hw_compat_2_11, hw_compat_2_11_len);
3332     vmc->smbios_old_sys_ver = true;
3333 }
3334 DEFINE_VIRT_MACHINE(2, 11)
3335 
3336 static void virt_machine_2_10_options(MachineClass *mc)
3337 {
3338     virt_machine_2_11_options(mc);
3339     compat_props_add(mc->compat_props, hw_compat_2_10, hw_compat_2_10_len);
3340     /* before 2.11 we never faulted accesses to bad addresses */
3341     mc->ignore_memory_transaction_failures = true;
3342 }
3343 DEFINE_VIRT_MACHINE(2, 10)
3344 
3345 static void virt_machine_2_9_options(MachineClass *mc)
3346 {
3347     virt_machine_2_10_options(mc);
3348     compat_props_add(mc->compat_props, hw_compat_2_9, hw_compat_2_9_len);
3349 }
3350 DEFINE_VIRT_MACHINE(2, 9)
3351 
3352 static void virt_machine_2_8_options(MachineClass *mc)
3353 {
3354     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3355 
3356     virt_machine_2_9_options(mc);
3357     compat_props_add(mc->compat_props, hw_compat_2_8, hw_compat_2_8_len);
3358     /* For 2.8 and earlier we falsely claimed in the DT that
3359      * our timers were edge-triggered, not level-triggered.
3360      */
3361     vmc->claim_edge_triggered_timers = true;
3362 }
3363 DEFINE_VIRT_MACHINE(2, 8)
3364 
3365 static void virt_machine_2_7_options(MachineClass *mc)
3366 {
3367     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3368 
3369     virt_machine_2_8_options(mc);
3370     compat_props_add(mc->compat_props, hw_compat_2_7, hw_compat_2_7_len);
3371     /* ITS was introduced with 2.8 */
3372     vmc->no_its = true;
3373     /* Stick with 1K pages for migration compatibility */
3374     mc->minimum_page_bits = 0;
3375 }
3376 DEFINE_VIRT_MACHINE(2, 7)
3377 
3378 static void virt_machine_2_6_options(MachineClass *mc)
3379 {
3380     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3381 
3382     virt_machine_2_7_options(mc);
3383     compat_props_add(mc->compat_props, hw_compat_2_6, hw_compat_2_6_len);
3384     vmc->disallow_affinity_adjustment = true;
3385     /* Disable PMU for 2.6 as PMU support was first introduced in 2.7 */
3386     vmc->no_pmu = true;
3387 }
3388 DEFINE_VIRT_MACHINE(2, 6)
3389