]> git.proxmox.com Git - mirror_qemu.git/blob - hw/arm/virt.c
Merge tag 'pull-ufs-20231030' of https://gitlab.com/jeuk20.kim/qemu into staging
[mirror_qemu.git] / hw / arm / virt.c
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("cortex-a710"),
215 ARM_CPU_TYPE_NAME("a64fx"),
216 ARM_CPU_TYPE_NAME("neoverse-n1"),
217 ARM_CPU_TYPE_NAME("neoverse-v1"),
218 ARM_CPU_TYPE_NAME("neoverse-n2"),
219 #endif
220 ARM_CPU_TYPE_NAME("cortex-a53"),
221 ARM_CPU_TYPE_NAME("cortex-a57"),
222 ARM_CPU_TYPE_NAME("host"),
223 ARM_CPU_TYPE_NAME("max"),
224 };
225
226 static bool cpu_type_valid(const char *cpu)
227 {
228 int i;
229
230 for (i = 0; i < ARRAY_SIZE(valid_cpus); i++) {
231 if (strcmp(cpu, valid_cpus[i]) == 0) {
232 return true;
233 }
234 }
235 return false;
236 }
237
238 static void create_randomness(MachineState *ms, const char *node)
239 {
240 struct {
241 uint64_t kaslr;
242 uint8_t rng[32];
243 } seed;
244
245 if (qemu_guest_getrandom(&seed, sizeof(seed), NULL)) {
246 return;
247 }
248 qemu_fdt_setprop_u64(ms->fdt, node, "kaslr-seed", seed.kaslr);
249 qemu_fdt_setprop(ms->fdt, node, "rng-seed", seed.rng, sizeof(seed.rng));
250 }
251
252 static void create_fdt(VirtMachineState *vms)
253 {
254 MachineState *ms = MACHINE(vms);
255 int nb_numa_nodes = ms->numa_state->num_nodes;
256 void *fdt = create_device_tree(&vms->fdt_size);
257
258 if (!fdt) {
259 error_report("create_device_tree() failed");
260 exit(1);
261 }
262
263 ms->fdt = fdt;
264
265 /* Header */
266 qemu_fdt_setprop_string(fdt, "/", "compatible", "linux,dummy-virt");
267 qemu_fdt_setprop_cell(fdt, "/", "#address-cells", 0x2);
268 qemu_fdt_setprop_cell(fdt, "/", "#size-cells", 0x2);
269 qemu_fdt_setprop_string(fdt, "/", "model", "linux,dummy-virt");
270
271 /* /chosen must exist for load_dtb to fill in necessary properties later */
272 qemu_fdt_add_subnode(fdt, "/chosen");
273 if (vms->dtb_randomness) {
274 create_randomness(ms, "/chosen");
275 }
276
277 if (vms->secure) {
278 qemu_fdt_add_subnode(fdt, "/secure-chosen");
279 if (vms->dtb_randomness) {
280 create_randomness(ms, "/secure-chosen");
281 }
282 }
283
284 /* Clock node, for the benefit of the UART. The kernel device tree
285 * binding documentation claims the PL011 node clock properties are
286 * optional but in practice if you omit them the kernel refuses to
287 * probe for the device.
288 */
289 vms->clock_phandle = qemu_fdt_alloc_phandle(fdt);
290 qemu_fdt_add_subnode(fdt, "/apb-pclk");
291 qemu_fdt_setprop_string(fdt, "/apb-pclk", "compatible", "fixed-clock");
292 qemu_fdt_setprop_cell(fdt, "/apb-pclk", "#clock-cells", 0x0);
293 qemu_fdt_setprop_cell(fdt, "/apb-pclk", "clock-frequency", 24000000);
294 qemu_fdt_setprop_string(fdt, "/apb-pclk", "clock-output-names",
295 "clk24mhz");
296 qemu_fdt_setprop_cell(fdt, "/apb-pclk", "phandle", vms->clock_phandle);
297
298 if (nb_numa_nodes > 0 && ms->numa_state->have_numa_distance) {
299 int size = nb_numa_nodes * nb_numa_nodes * 3 * sizeof(uint32_t);
300 uint32_t *matrix = g_malloc0(size);
301 int idx, i, j;
302
303 for (i = 0; i < nb_numa_nodes; i++) {
304 for (j = 0; j < nb_numa_nodes; j++) {
305 idx = (i * nb_numa_nodes + j) * 3;
306 matrix[idx + 0] = cpu_to_be32(i);
307 matrix[idx + 1] = cpu_to_be32(j);
308 matrix[idx + 2] =
309 cpu_to_be32(ms->numa_state->nodes[i].distance[j]);
310 }
311 }
312
313 qemu_fdt_add_subnode(fdt, "/distance-map");
314 qemu_fdt_setprop_string(fdt, "/distance-map", "compatible",
315 "numa-distance-map-v1");
316 qemu_fdt_setprop(fdt, "/distance-map", "distance-matrix",
317 matrix, size);
318 g_free(matrix);
319 }
320 }
321
322 static void fdt_add_timer_nodes(const VirtMachineState *vms)
323 {
324 /* On real hardware these interrupts are level-triggered.
325 * On KVM they were edge-triggered before host kernel version 4.4,
326 * and level-triggered afterwards.
327 * On emulated QEMU they are level-triggered.
328 *
329 * Getting the DTB info about them wrong is awkward for some
330 * guest kernels:
331 * pre-4.8 ignore the DT and leave the interrupt configured
332 * with whatever the GIC reset value (or the bootloader) left it at
333 * 4.8 before rc6 honour the incorrect data by programming it back
334 * into the GIC, causing problems
335 * 4.8rc6 and later ignore the DT and always write "level triggered"
336 * into the GIC
337 *
338 * For backwards-compatibility, virt-2.8 and earlier will continue
339 * to say these are edge-triggered, but later machines will report
340 * the correct information.
341 */
342 ARMCPU *armcpu;
343 VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
344 uint32_t irqflags = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
345 MachineState *ms = MACHINE(vms);
346
347 if (vmc->claim_edge_triggered_timers) {
348 irqflags = GIC_FDT_IRQ_FLAGS_EDGE_LO_HI;
349 }
350
351 if (vms->gic_version == VIRT_GIC_VERSION_2) {
352 irqflags = deposit32(irqflags, GIC_FDT_IRQ_PPI_CPU_START,
353 GIC_FDT_IRQ_PPI_CPU_WIDTH,
354 (1 << MACHINE(vms)->smp.cpus) - 1);
355 }
356
357 qemu_fdt_add_subnode(ms->fdt, "/timer");
358
359 armcpu = ARM_CPU(qemu_get_cpu(0));
360 if (arm_feature(&armcpu->env, ARM_FEATURE_V8)) {
361 const char compat[] = "arm,armv8-timer\0arm,armv7-timer";
362 qemu_fdt_setprop(ms->fdt, "/timer", "compatible",
363 compat, sizeof(compat));
364 } else {
365 qemu_fdt_setprop_string(ms->fdt, "/timer", "compatible",
366 "arm,armv7-timer");
367 }
368 qemu_fdt_setprop(ms->fdt, "/timer", "always-on", NULL, 0);
369 qemu_fdt_setprop_cells(ms->fdt, "/timer", "interrupts",
370 GIC_FDT_IRQ_TYPE_PPI,
371 INTID_TO_PPI(ARCH_TIMER_S_EL1_IRQ), irqflags,
372 GIC_FDT_IRQ_TYPE_PPI,
373 INTID_TO_PPI(ARCH_TIMER_NS_EL1_IRQ), irqflags,
374 GIC_FDT_IRQ_TYPE_PPI,
375 INTID_TO_PPI(ARCH_TIMER_VIRT_IRQ), irqflags,
376 GIC_FDT_IRQ_TYPE_PPI,
377 INTID_TO_PPI(ARCH_TIMER_NS_EL2_IRQ), irqflags);
378 }
379
380 static void fdt_add_cpu_nodes(const VirtMachineState *vms)
381 {
382 int cpu;
383 int addr_cells = 1;
384 const MachineState *ms = MACHINE(vms);
385 const VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
386 int smp_cpus = ms->smp.cpus;
387
388 /*
389 * See Linux Documentation/devicetree/bindings/arm/cpus.yaml
390 * On ARM v8 64-bit systems value should be set to 2,
391 * that corresponds to the MPIDR_EL1 register size.
392 * If MPIDR_EL1[63:32] value is equal to 0 on all CPUs
393 * in the system, #address-cells can be set to 1, since
394 * MPIDR_EL1[63:32] bits are not used for CPUs
395 * identification.
396 *
397 * Here we actually don't know whether our system is 32- or 64-bit one.
398 * The simplest way to go is to examine affinity IDs of all our CPUs. If
399 * at least one of them has Aff3 populated, we set #address-cells to 2.
400 */
401 for (cpu = 0; cpu < smp_cpus; cpu++) {
402 ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
403
404 if (armcpu->mp_affinity & ARM_AFF3_MASK) {
405 addr_cells = 2;
406 break;
407 }
408 }
409
410 qemu_fdt_add_subnode(ms->fdt, "/cpus");
411 qemu_fdt_setprop_cell(ms->fdt, "/cpus", "#address-cells", addr_cells);
412 qemu_fdt_setprop_cell(ms->fdt, "/cpus", "#size-cells", 0x0);
413
414 for (cpu = smp_cpus - 1; cpu >= 0; cpu--) {
415 char *nodename = g_strdup_printf("/cpus/cpu@%d", cpu);
416 ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
417 CPUState *cs = CPU(armcpu);
418
419 qemu_fdt_add_subnode(ms->fdt, nodename);
420 qemu_fdt_setprop_string(ms->fdt, nodename, "device_type", "cpu");
421 qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
422 armcpu->dtb_compatible);
423
424 if (vms->psci_conduit != QEMU_PSCI_CONDUIT_DISABLED && smp_cpus > 1) {
425 qemu_fdt_setprop_string(ms->fdt, nodename,
426 "enable-method", "psci");
427 }
428
429 if (addr_cells == 2) {
430 qemu_fdt_setprop_u64(ms->fdt, nodename, "reg",
431 armcpu->mp_affinity);
432 } else {
433 qemu_fdt_setprop_cell(ms->fdt, nodename, "reg",
434 armcpu->mp_affinity);
435 }
436
437 if (ms->possible_cpus->cpus[cs->cpu_index].props.has_node_id) {
438 qemu_fdt_setprop_cell(ms->fdt, nodename, "numa-node-id",
439 ms->possible_cpus->cpus[cs->cpu_index].props.node_id);
440 }
441
442 if (!vmc->no_cpu_topology) {
443 qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle",
444 qemu_fdt_alloc_phandle(ms->fdt));
445 }
446
447 g_free(nodename);
448 }
449
450 if (!vmc->no_cpu_topology) {
451 /*
452 * Add vCPU topology description through fdt node cpu-map.
453 *
454 * See Linux Documentation/devicetree/bindings/cpu/cpu-topology.txt
455 * In a SMP system, the hierarchy of CPUs can be defined through
456 * four entities that are used to describe the layout of CPUs in
457 * the system: socket/cluster/core/thread.
458 *
459 * A socket node represents the boundary of system physical package
460 * and its child nodes must be one or more cluster nodes. A system
461 * can contain several layers of clustering within a single physical
462 * package and cluster nodes can be contained in parent cluster nodes.
463 *
464 * Note: currently we only support one layer of clustering within
465 * each physical package.
466 */
467 qemu_fdt_add_subnode(ms->fdt, "/cpus/cpu-map");
468
469 for (cpu = smp_cpus - 1; cpu >= 0; cpu--) {
470 char *cpu_path = g_strdup_printf("/cpus/cpu@%d", cpu);
471 char *map_path;
472
473 if (ms->smp.threads > 1) {
474 map_path = g_strdup_printf(
475 "/cpus/cpu-map/socket%d/cluster%d/core%d/thread%d",
476 cpu / (ms->smp.clusters * ms->smp.cores * ms->smp.threads),
477 (cpu / (ms->smp.cores * ms->smp.threads)) % ms->smp.clusters,
478 (cpu / ms->smp.threads) % ms->smp.cores,
479 cpu % ms->smp.threads);
480 } else {
481 map_path = g_strdup_printf(
482 "/cpus/cpu-map/socket%d/cluster%d/core%d",
483 cpu / (ms->smp.clusters * ms->smp.cores),
484 (cpu / ms->smp.cores) % ms->smp.clusters,
485 cpu % ms->smp.cores);
486 }
487 qemu_fdt_add_path(ms->fdt, map_path);
488 qemu_fdt_setprop_phandle(ms->fdt, map_path, "cpu", cpu_path);
489
490 g_free(map_path);
491 g_free(cpu_path);
492 }
493 }
494 }
495
496 static void fdt_add_its_gic_node(VirtMachineState *vms)
497 {
498 char *nodename;
499 MachineState *ms = MACHINE(vms);
500
501 vms->msi_phandle = qemu_fdt_alloc_phandle(ms->fdt);
502 nodename = g_strdup_printf("/intc/its@%" PRIx64,
503 vms->memmap[VIRT_GIC_ITS].base);
504 qemu_fdt_add_subnode(ms->fdt, nodename);
505 qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
506 "arm,gic-v3-its");
507 qemu_fdt_setprop(ms->fdt, nodename, "msi-controller", NULL, 0);
508 qemu_fdt_setprop_cell(ms->fdt, nodename, "#msi-cells", 1);
509 qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
510 2, vms->memmap[VIRT_GIC_ITS].base,
511 2, vms->memmap[VIRT_GIC_ITS].size);
512 qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", vms->msi_phandle);
513 g_free(nodename);
514 }
515
516 static void fdt_add_v2m_gic_node(VirtMachineState *vms)
517 {
518 MachineState *ms = MACHINE(vms);
519 char *nodename;
520
521 nodename = g_strdup_printf("/intc/v2m@%" PRIx64,
522 vms->memmap[VIRT_GIC_V2M].base);
523 vms->msi_phandle = qemu_fdt_alloc_phandle(ms->fdt);
524 qemu_fdt_add_subnode(ms->fdt, nodename);
525 qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
526 "arm,gic-v2m-frame");
527 qemu_fdt_setprop(ms->fdt, nodename, "msi-controller", NULL, 0);
528 qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
529 2, vms->memmap[VIRT_GIC_V2M].base,
530 2, vms->memmap[VIRT_GIC_V2M].size);
531 qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", vms->msi_phandle);
532 g_free(nodename);
533 }
534
535 static void fdt_add_gic_node(VirtMachineState *vms)
536 {
537 MachineState *ms = MACHINE(vms);
538 char *nodename;
539
540 vms->gic_phandle = qemu_fdt_alloc_phandle(ms->fdt);
541 qemu_fdt_setprop_cell(ms->fdt, "/", "interrupt-parent", vms->gic_phandle);
542
543 nodename = g_strdup_printf("/intc@%" PRIx64,
544 vms->memmap[VIRT_GIC_DIST].base);
545 qemu_fdt_add_subnode(ms->fdt, nodename);
546 qemu_fdt_setprop_cell(ms->fdt, nodename, "#interrupt-cells", 3);
547 qemu_fdt_setprop(ms->fdt, nodename, "interrupt-controller", NULL, 0);
548 qemu_fdt_setprop_cell(ms->fdt, nodename, "#address-cells", 0x2);
549 qemu_fdt_setprop_cell(ms->fdt, nodename, "#size-cells", 0x2);
550 qemu_fdt_setprop(ms->fdt, nodename, "ranges", NULL, 0);
551 if (vms->gic_version != VIRT_GIC_VERSION_2) {
552 int nb_redist_regions = virt_gicv3_redist_region_count(vms);
553
554 qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
555 "arm,gic-v3");
556
557 qemu_fdt_setprop_cell(ms->fdt, nodename,
558 "#redistributor-regions", nb_redist_regions);
559
560 if (nb_redist_regions == 1) {
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 } else {
567 qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
568 2, vms->memmap[VIRT_GIC_DIST].base,
569 2, vms->memmap[VIRT_GIC_DIST].size,
570 2, vms->memmap[VIRT_GIC_REDIST].base,
571 2, vms->memmap[VIRT_GIC_REDIST].size,
572 2, vms->memmap[VIRT_HIGH_GIC_REDIST2].base,
573 2, vms->memmap[VIRT_HIGH_GIC_REDIST2].size);
574 }
575
576 if (vms->virt) {
577 qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
578 GIC_FDT_IRQ_TYPE_PPI, ARCH_GIC_MAINT_IRQ,
579 GIC_FDT_IRQ_FLAGS_LEVEL_HI);
580 }
581 } else {
582 /* 'cortex-a15-gic' means 'GIC v2' */
583 qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
584 "arm,cortex-a15-gic");
585 if (!vms->virt) {
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 } else {
592 qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
593 2, vms->memmap[VIRT_GIC_DIST].base,
594 2, vms->memmap[VIRT_GIC_DIST].size,
595 2, vms->memmap[VIRT_GIC_CPU].base,
596 2, vms->memmap[VIRT_GIC_CPU].size,
597 2, vms->memmap[VIRT_GIC_HYP].base,
598 2, vms->memmap[VIRT_GIC_HYP].size,
599 2, vms->memmap[VIRT_GIC_VCPU].base,
600 2, vms->memmap[VIRT_GIC_VCPU].size);
601 qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
602 GIC_FDT_IRQ_TYPE_PPI, ARCH_GIC_MAINT_IRQ,
603 GIC_FDT_IRQ_FLAGS_LEVEL_HI);
604 }
605 }
606
607 qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", vms->gic_phandle);
608 g_free(nodename);
609 }
610
611 static void fdt_add_pmu_nodes(const VirtMachineState *vms)
612 {
613 ARMCPU *armcpu = ARM_CPU(first_cpu);
614 uint32_t irqflags = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
615 MachineState *ms = MACHINE(vms);
616
617 if (!arm_feature(&armcpu->env, ARM_FEATURE_PMU)) {
618 assert(!object_property_get_bool(OBJECT(armcpu), "pmu", NULL));
619 return;
620 }
621
622 if (vms->gic_version == VIRT_GIC_VERSION_2) {
623 irqflags = deposit32(irqflags, GIC_FDT_IRQ_PPI_CPU_START,
624 GIC_FDT_IRQ_PPI_CPU_WIDTH,
625 (1 << MACHINE(vms)->smp.cpus) - 1);
626 }
627
628 qemu_fdt_add_subnode(ms->fdt, "/pmu");
629 if (arm_feature(&armcpu->env, ARM_FEATURE_V8)) {
630 const char compat[] = "arm,armv8-pmuv3";
631 qemu_fdt_setprop(ms->fdt, "/pmu", "compatible",
632 compat, sizeof(compat));
633 qemu_fdt_setprop_cells(ms->fdt, "/pmu", "interrupts",
634 GIC_FDT_IRQ_TYPE_PPI, VIRTUAL_PMU_IRQ, irqflags);
635 }
636 }
637
638 static inline DeviceState *create_acpi_ged(VirtMachineState *vms)
639 {
640 DeviceState *dev;
641 MachineState *ms = MACHINE(vms);
642 int irq = vms->irqmap[VIRT_ACPI_GED];
643 uint32_t event = ACPI_GED_PWR_DOWN_EVT;
644
645 if (ms->ram_slots) {
646 event |= ACPI_GED_MEM_HOTPLUG_EVT;
647 }
648
649 if (ms->nvdimms_state->is_enabled) {
650 event |= ACPI_GED_NVDIMM_HOTPLUG_EVT;
651 }
652
653 dev = qdev_new(TYPE_ACPI_GED);
654 qdev_prop_set_uint32(dev, "ged-event", event);
655 sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
656
657 sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_ACPI_GED].base);
658 sysbus_mmio_map(SYS_BUS_DEVICE(dev), 1, vms->memmap[VIRT_PCDIMM_ACPI].base);
659 sysbus_connect_irq(SYS_BUS_DEVICE(dev), 0, qdev_get_gpio_in(vms->gic, irq));
660
661 return dev;
662 }
663
664 static void create_its(VirtMachineState *vms)
665 {
666 const char *itsclass = its_class_name();
667 DeviceState *dev;
668
669 if (!strcmp(itsclass, "arm-gicv3-its")) {
670 if (!vms->tcg_its) {
671 itsclass = NULL;
672 }
673 }
674
675 if (!itsclass) {
676 /* Do nothing if not supported */
677 return;
678 }
679
680 dev = qdev_new(itsclass);
681
682 object_property_set_link(OBJECT(dev), "parent-gicv3", OBJECT(vms->gic),
683 &error_abort);
684 sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
685 sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_ITS].base);
686
687 fdt_add_its_gic_node(vms);
688 vms->msi_controller = VIRT_MSI_CTRL_ITS;
689 }
690
691 static void create_v2m(VirtMachineState *vms)
692 {
693 int i;
694 int irq = vms->irqmap[VIRT_GIC_V2M];
695 DeviceState *dev;
696
697 dev = qdev_new("arm-gicv2m");
698 qdev_prop_set_uint32(dev, "base-spi", irq);
699 qdev_prop_set_uint32(dev, "num-spi", NUM_GICV2M_SPIS);
700 sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
701 sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_V2M].base);
702
703 for (i = 0; i < NUM_GICV2M_SPIS; i++) {
704 sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
705 qdev_get_gpio_in(vms->gic, irq + i));
706 }
707
708 fdt_add_v2m_gic_node(vms);
709 vms->msi_controller = VIRT_MSI_CTRL_GICV2M;
710 }
711
712 static void create_gic(VirtMachineState *vms, MemoryRegion *mem)
713 {
714 MachineState *ms = MACHINE(vms);
715 /* We create a standalone GIC */
716 SysBusDevice *gicbusdev;
717 const char *gictype;
718 int i;
719 unsigned int smp_cpus = ms->smp.cpus;
720 uint32_t nb_redist_regions = 0;
721 int revision;
722
723 if (vms->gic_version == VIRT_GIC_VERSION_2) {
724 gictype = gic_class_name();
725 } else {
726 gictype = gicv3_class_name();
727 }
728
729 switch (vms->gic_version) {
730 case VIRT_GIC_VERSION_2:
731 revision = 2;
732 break;
733 case VIRT_GIC_VERSION_3:
734 revision = 3;
735 break;
736 case VIRT_GIC_VERSION_4:
737 revision = 4;
738 break;
739 default:
740 g_assert_not_reached();
741 }
742 vms->gic = qdev_new(gictype);
743 qdev_prop_set_uint32(vms->gic, "revision", revision);
744 qdev_prop_set_uint32(vms->gic, "num-cpu", smp_cpus);
745 /* Note that the num-irq property counts both internal and external
746 * interrupts; there are always 32 of the former (mandated by GIC spec).
747 */
748 qdev_prop_set_uint32(vms->gic, "num-irq", NUM_IRQS + 32);
749 if (!kvm_irqchip_in_kernel()) {
750 qdev_prop_set_bit(vms->gic, "has-security-extensions", vms->secure);
751 }
752
753 if (vms->gic_version != VIRT_GIC_VERSION_2) {
754 uint32_t redist0_capacity = virt_redist_capacity(vms, VIRT_GIC_REDIST);
755 uint32_t redist0_count = MIN(smp_cpus, redist0_capacity);
756
757 nb_redist_regions = virt_gicv3_redist_region_count(vms);
758
759 qdev_prop_set_uint32(vms->gic, "len-redist-region-count",
760 nb_redist_regions);
761 qdev_prop_set_uint32(vms->gic, "redist-region-count[0]", redist0_count);
762
763 if (!kvm_irqchip_in_kernel()) {
764 if (vms->tcg_its) {
765 object_property_set_link(OBJECT(vms->gic), "sysmem",
766 OBJECT(mem), &error_fatal);
767 qdev_prop_set_bit(vms->gic, "has-lpi", true);
768 }
769 }
770
771 if (nb_redist_regions == 2) {
772 uint32_t redist1_capacity =
773 virt_redist_capacity(vms, VIRT_HIGH_GIC_REDIST2);
774
775 qdev_prop_set_uint32(vms->gic, "redist-region-count[1]",
776 MIN(smp_cpus - redist0_count, redist1_capacity));
777 }
778 } else {
779 if (!kvm_irqchip_in_kernel()) {
780 qdev_prop_set_bit(vms->gic, "has-virtualization-extensions",
781 vms->virt);
782 }
783 }
784 gicbusdev = SYS_BUS_DEVICE(vms->gic);
785 sysbus_realize_and_unref(gicbusdev, &error_fatal);
786 sysbus_mmio_map(gicbusdev, 0, vms->memmap[VIRT_GIC_DIST].base);
787 if (vms->gic_version != VIRT_GIC_VERSION_2) {
788 sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_REDIST].base);
789 if (nb_redist_regions == 2) {
790 sysbus_mmio_map(gicbusdev, 2,
791 vms->memmap[VIRT_HIGH_GIC_REDIST2].base);
792 }
793 } else {
794 sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_CPU].base);
795 if (vms->virt) {
796 sysbus_mmio_map(gicbusdev, 2, vms->memmap[VIRT_GIC_HYP].base);
797 sysbus_mmio_map(gicbusdev, 3, vms->memmap[VIRT_GIC_VCPU].base);
798 }
799 }
800
801 /* Wire the outputs from each CPU's generic timer and the GICv3
802 * maintenance interrupt signal to the appropriate GIC PPI inputs,
803 * and the GIC's IRQ/FIQ/VIRQ/VFIQ interrupt outputs to the CPU's inputs.
804 */
805 for (i = 0; i < smp_cpus; i++) {
806 DeviceState *cpudev = DEVICE(qemu_get_cpu(i));
807 int intidbase = NUM_IRQS + i * GIC_INTERNAL;
808 /* Mapping from the output timer irq lines from the CPU to the
809 * GIC PPI inputs we use for the virt board.
810 */
811 const int timer_irq[] = {
812 [GTIMER_PHYS] = ARCH_TIMER_NS_EL1_IRQ,
813 [GTIMER_VIRT] = ARCH_TIMER_VIRT_IRQ,
814 [GTIMER_HYP] = ARCH_TIMER_NS_EL2_IRQ,
815 [GTIMER_SEC] = ARCH_TIMER_S_EL1_IRQ,
816 };
817
818 for (unsigned irq = 0; irq < ARRAY_SIZE(timer_irq); irq++) {
819 qdev_connect_gpio_out(cpudev, irq,
820 qdev_get_gpio_in(vms->gic,
821 intidbase + timer_irq[irq]));
822 }
823
824 if (vms->gic_version != VIRT_GIC_VERSION_2) {
825 qemu_irq irq = qdev_get_gpio_in(vms->gic,
826 intidbase + ARCH_GIC_MAINT_IRQ);
827 qdev_connect_gpio_out_named(cpudev, "gicv3-maintenance-interrupt",
828 0, irq);
829 } else if (vms->virt) {
830 qemu_irq irq = qdev_get_gpio_in(vms->gic,
831 intidbase + ARCH_GIC_MAINT_IRQ);
832 sysbus_connect_irq(gicbusdev, i + 4 * smp_cpus, irq);
833 }
834
835 qdev_connect_gpio_out_named(cpudev, "pmu-interrupt", 0,
836 qdev_get_gpio_in(vms->gic, intidbase
837 + VIRTUAL_PMU_IRQ));
838
839 sysbus_connect_irq(gicbusdev, i, qdev_get_gpio_in(cpudev, ARM_CPU_IRQ));
840 sysbus_connect_irq(gicbusdev, i + smp_cpus,
841 qdev_get_gpio_in(cpudev, ARM_CPU_FIQ));
842 sysbus_connect_irq(gicbusdev, i + 2 * smp_cpus,
843 qdev_get_gpio_in(cpudev, ARM_CPU_VIRQ));
844 sysbus_connect_irq(gicbusdev, i + 3 * smp_cpus,
845 qdev_get_gpio_in(cpudev, ARM_CPU_VFIQ));
846 }
847
848 fdt_add_gic_node(vms);
849
850 if (vms->gic_version != VIRT_GIC_VERSION_2 && vms->its) {
851 create_its(vms);
852 } else if (vms->gic_version == VIRT_GIC_VERSION_2) {
853 create_v2m(vms);
854 }
855 }
856
857 static void create_uart(const VirtMachineState *vms, int uart,
858 MemoryRegion *mem, Chardev *chr)
859 {
860 char *nodename;
861 hwaddr base = vms->memmap[uart].base;
862 hwaddr size = vms->memmap[uart].size;
863 int irq = vms->irqmap[uart];
864 const char compat[] = "arm,pl011\0arm,primecell";
865 const char clocknames[] = "uartclk\0apb_pclk";
866 DeviceState *dev = qdev_new(TYPE_PL011);
867 SysBusDevice *s = SYS_BUS_DEVICE(dev);
868 MachineState *ms = MACHINE(vms);
869
870 qdev_prop_set_chr(dev, "chardev", chr);
871 sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
872 memory_region_add_subregion(mem, base,
873 sysbus_mmio_get_region(s, 0));
874 sysbus_connect_irq(s, 0, qdev_get_gpio_in(vms->gic, irq));
875
876 nodename = g_strdup_printf("/pl011@%" PRIx64, base);
877 qemu_fdt_add_subnode(ms->fdt, nodename);
878 /* Note that we can't use setprop_string because of the embedded NUL */
879 qemu_fdt_setprop(ms->fdt, nodename, "compatible",
880 compat, sizeof(compat));
881 qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
882 2, base, 2, size);
883 qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
884 GIC_FDT_IRQ_TYPE_SPI, irq,
885 GIC_FDT_IRQ_FLAGS_LEVEL_HI);
886 qemu_fdt_setprop_cells(ms->fdt, nodename, "clocks",
887 vms->clock_phandle, vms->clock_phandle);
888 qemu_fdt_setprop(ms->fdt, nodename, "clock-names",
889 clocknames, sizeof(clocknames));
890
891 if (uart == VIRT_UART) {
892 qemu_fdt_setprop_string(ms->fdt, "/chosen", "stdout-path", nodename);
893 } else {
894 /* Mark as not usable by the normal world */
895 qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
896 qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
897
898 qemu_fdt_setprop_string(ms->fdt, "/secure-chosen", "stdout-path",
899 nodename);
900 }
901
902 g_free(nodename);
903 }
904
905 static void create_rtc(const VirtMachineState *vms)
906 {
907 char *nodename;
908 hwaddr base = vms->memmap[VIRT_RTC].base;
909 hwaddr size = vms->memmap[VIRT_RTC].size;
910 int irq = vms->irqmap[VIRT_RTC];
911 const char compat[] = "arm,pl031\0arm,primecell";
912 MachineState *ms = MACHINE(vms);
913
914 sysbus_create_simple("pl031", base, qdev_get_gpio_in(vms->gic, irq));
915
916 nodename = g_strdup_printf("/pl031@%" PRIx64, base);
917 qemu_fdt_add_subnode(ms->fdt, nodename);
918 qemu_fdt_setprop(ms->fdt, nodename, "compatible", compat, sizeof(compat));
919 qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
920 2, base, 2, size);
921 qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
922 GIC_FDT_IRQ_TYPE_SPI, irq,
923 GIC_FDT_IRQ_FLAGS_LEVEL_HI);
924 qemu_fdt_setprop_cell(ms->fdt, nodename, "clocks", vms->clock_phandle);
925 qemu_fdt_setprop_string(ms->fdt, nodename, "clock-names", "apb_pclk");
926 g_free(nodename);
927 }
928
929 static DeviceState *gpio_key_dev;
930 static void virt_powerdown_req(Notifier *n, void *opaque)
931 {
932 VirtMachineState *s = container_of(n, VirtMachineState, powerdown_notifier);
933
934 if (s->acpi_dev) {
935 acpi_send_event(s->acpi_dev, ACPI_POWER_DOWN_STATUS);
936 } else {
937 /* use gpio Pin 3 for power button event */
938 qemu_set_irq(qdev_get_gpio_in(gpio_key_dev, 0), 1);
939 }
940 }
941
942 static void create_gpio_keys(char *fdt, DeviceState *pl061_dev,
943 uint32_t phandle)
944 {
945 gpio_key_dev = sysbus_create_simple("gpio-key", -1,
946 qdev_get_gpio_in(pl061_dev, 3));
947
948 qemu_fdt_add_subnode(fdt, "/gpio-keys");
949 qemu_fdt_setprop_string(fdt, "/gpio-keys", "compatible", "gpio-keys");
950
951 qemu_fdt_add_subnode(fdt, "/gpio-keys/poweroff");
952 qemu_fdt_setprop_string(fdt, "/gpio-keys/poweroff",
953 "label", "GPIO Key Poweroff");
954 qemu_fdt_setprop_cell(fdt, "/gpio-keys/poweroff", "linux,code",
955 KEY_POWER);
956 qemu_fdt_setprop_cells(fdt, "/gpio-keys/poweroff",
957 "gpios", phandle, 3, 0);
958 }
959
960 #define SECURE_GPIO_POWEROFF 0
961 #define SECURE_GPIO_RESET 1
962
963 static void create_secure_gpio_pwr(char *fdt, DeviceState *pl061_dev,
964 uint32_t phandle)
965 {
966 DeviceState *gpio_pwr_dev;
967
968 /* gpio-pwr */
969 gpio_pwr_dev = sysbus_create_simple("gpio-pwr", -1, NULL);
970
971 /* connect secure pl061 to gpio-pwr */
972 qdev_connect_gpio_out(pl061_dev, SECURE_GPIO_RESET,
973 qdev_get_gpio_in_named(gpio_pwr_dev, "reset", 0));
974 qdev_connect_gpio_out(pl061_dev, SECURE_GPIO_POWEROFF,
975 qdev_get_gpio_in_named(gpio_pwr_dev, "shutdown", 0));
976
977 qemu_fdt_add_subnode(fdt, "/gpio-poweroff");
978 qemu_fdt_setprop_string(fdt, "/gpio-poweroff", "compatible",
979 "gpio-poweroff");
980 qemu_fdt_setprop_cells(fdt, "/gpio-poweroff",
981 "gpios", phandle, SECURE_GPIO_POWEROFF, 0);
982 qemu_fdt_setprop_string(fdt, "/gpio-poweroff", "status", "disabled");
983 qemu_fdt_setprop_string(fdt, "/gpio-poweroff", "secure-status",
984 "okay");
985
986 qemu_fdt_add_subnode(fdt, "/gpio-restart");
987 qemu_fdt_setprop_string(fdt, "/gpio-restart", "compatible",
988 "gpio-restart");
989 qemu_fdt_setprop_cells(fdt, "/gpio-restart",
990 "gpios", phandle, SECURE_GPIO_RESET, 0);
991 qemu_fdt_setprop_string(fdt, "/gpio-restart", "status", "disabled");
992 qemu_fdt_setprop_string(fdt, "/gpio-restart", "secure-status",
993 "okay");
994 }
995
996 static void create_gpio_devices(const VirtMachineState *vms, int gpio,
997 MemoryRegion *mem)
998 {
999 char *nodename;
1000 DeviceState *pl061_dev;
1001 hwaddr base = vms->memmap[gpio].base;
1002 hwaddr size = vms->memmap[gpio].size;
1003 int irq = vms->irqmap[gpio];
1004 const char compat[] = "arm,pl061\0arm,primecell";
1005 SysBusDevice *s;
1006 MachineState *ms = MACHINE(vms);
1007
1008 pl061_dev = qdev_new("pl061");
1009 /* Pull lines down to 0 if not driven by the PL061 */
1010 qdev_prop_set_uint32(pl061_dev, "pullups", 0);
1011 qdev_prop_set_uint32(pl061_dev, "pulldowns", 0xff);
1012 s = SYS_BUS_DEVICE(pl061_dev);
1013 sysbus_realize_and_unref(s, &error_fatal);
1014 memory_region_add_subregion(mem, base, sysbus_mmio_get_region(s, 0));
1015 sysbus_connect_irq(s, 0, qdev_get_gpio_in(vms->gic, irq));
1016
1017 uint32_t phandle = qemu_fdt_alloc_phandle(ms->fdt);
1018 nodename = g_strdup_printf("/pl061@%" PRIx64, base);
1019 qemu_fdt_add_subnode(ms->fdt, nodename);
1020 qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1021 2, base, 2, size);
1022 qemu_fdt_setprop(ms->fdt, nodename, "compatible", compat, sizeof(compat));
1023 qemu_fdt_setprop_cell(ms->fdt, nodename, "#gpio-cells", 2);
1024 qemu_fdt_setprop(ms->fdt, nodename, "gpio-controller", NULL, 0);
1025 qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
1026 GIC_FDT_IRQ_TYPE_SPI, irq,
1027 GIC_FDT_IRQ_FLAGS_LEVEL_HI);
1028 qemu_fdt_setprop_cell(ms->fdt, nodename, "clocks", vms->clock_phandle);
1029 qemu_fdt_setprop_string(ms->fdt, nodename, "clock-names", "apb_pclk");
1030 qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", phandle);
1031
1032 if (gpio != VIRT_GPIO) {
1033 /* Mark as not usable by the normal world */
1034 qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
1035 qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
1036 }
1037 g_free(nodename);
1038
1039 /* Child gpio devices */
1040 if (gpio == VIRT_GPIO) {
1041 create_gpio_keys(ms->fdt, pl061_dev, phandle);
1042 } else {
1043 create_secure_gpio_pwr(ms->fdt, pl061_dev, phandle);
1044 }
1045 }
1046
1047 static void create_virtio_devices(const VirtMachineState *vms)
1048 {
1049 int i;
1050 hwaddr size = vms->memmap[VIRT_MMIO].size;
1051 MachineState *ms = MACHINE(vms);
1052
1053 /* We create the transports in forwards order. Since qbus_realize()
1054 * prepends (not appends) new child buses, the incrementing loop below will
1055 * create a list of virtio-mmio buses with decreasing base addresses.
1056 *
1057 * When a -device option is processed from the command line,
1058 * qbus_find_recursive() picks the next free virtio-mmio bus in forwards
1059 * order. The upshot is that -device options in increasing command line
1060 * order are mapped to virtio-mmio buses with decreasing base addresses.
1061 *
1062 * When this code was originally written, that arrangement ensured that the
1063 * guest Linux kernel would give the lowest "name" (/dev/vda, eth0, etc) to
1064 * the first -device on the command line. (The end-to-end order is a
1065 * function of this loop, qbus_realize(), qbus_find_recursive(), and the
1066 * guest kernel's name-to-address assignment strategy.)
1067 *
1068 * Meanwhile, the kernel's traversal seems to have been reversed; see eg.
1069 * the message, if not necessarily the code, of commit 70161ff336.
1070 * Therefore the loop now establishes the inverse of the original intent.
1071 *
1072 * Unfortunately, we can't counteract the kernel change by reversing the
1073 * loop; it would break existing command lines.
1074 *
1075 * In any case, the kernel makes no guarantee about the stability of
1076 * enumeration order of virtio devices (as demonstrated by it changing
1077 * between kernel versions). For reliable and stable identification
1078 * of disks users must use UUIDs or similar mechanisms.
1079 */
1080 for (i = 0; i < NUM_VIRTIO_TRANSPORTS; i++) {
1081 int irq = vms->irqmap[VIRT_MMIO] + i;
1082 hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
1083
1084 sysbus_create_simple("virtio-mmio", base,
1085 qdev_get_gpio_in(vms->gic, irq));
1086 }
1087
1088 /* We add dtb nodes in reverse order so that they appear in the finished
1089 * device tree lowest address first.
1090 *
1091 * Note that this mapping is independent of the loop above. The previous
1092 * loop influences virtio device to virtio transport assignment, whereas
1093 * this loop controls how virtio transports are laid out in the dtb.
1094 */
1095 for (i = NUM_VIRTIO_TRANSPORTS - 1; i >= 0; i--) {
1096 char *nodename;
1097 int irq = vms->irqmap[VIRT_MMIO] + i;
1098 hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
1099
1100 nodename = g_strdup_printf("/virtio_mmio@%" PRIx64, base);
1101 qemu_fdt_add_subnode(ms->fdt, nodename);
1102 qemu_fdt_setprop_string(ms->fdt, nodename,
1103 "compatible", "virtio,mmio");
1104 qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1105 2, base, 2, size);
1106 qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
1107 GIC_FDT_IRQ_TYPE_SPI, irq,
1108 GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
1109 qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1110 g_free(nodename);
1111 }
1112 }
1113
1114 #define VIRT_FLASH_SECTOR_SIZE (256 * KiB)
1115
1116 static PFlashCFI01 *virt_flash_create1(VirtMachineState *vms,
1117 const char *name,
1118 const char *alias_prop_name)
1119 {
1120 /*
1121 * Create a single flash device. We use the same parameters as
1122 * the flash devices on the Versatile Express board.
1123 */
1124 DeviceState *dev = qdev_new(TYPE_PFLASH_CFI01);
1125
1126 qdev_prop_set_uint64(dev, "sector-length", VIRT_FLASH_SECTOR_SIZE);
1127 qdev_prop_set_uint8(dev, "width", 4);
1128 qdev_prop_set_uint8(dev, "device-width", 2);
1129 qdev_prop_set_bit(dev, "big-endian", false);
1130 qdev_prop_set_uint16(dev, "id0", 0x89);
1131 qdev_prop_set_uint16(dev, "id1", 0x18);
1132 qdev_prop_set_uint16(dev, "id2", 0x00);
1133 qdev_prop_set_uint16(dev, "id3", 0x00);
1134 qdev_prop_set_string(dev, "name", name);
1135 object_property_add_child(OBJECT(vms), name, OBJECT(dev));
1136 object_property_add_alias(OBJECT(vms), alias_prop_name,
1137 OBJECT(dev), "drive");
1138 return PFLASH_CFI01(dev);
1139 }
1140
1141 static void virt_flash_create(VirtMachineState *vms)
1142 {
1143 vms->flash[0] = virt_flash_create1(vms, "virt.flash0", "pflash0");
1144 vms->flash[1] = virt_flash_create1(vms, "virt.flash1", "pflash1");
1145 }
1146
1147 static void virt_flash_map1(PFlashCFI01 *flash,
1148 hwaddr base, hwaddr size,
1149 MemoryRegion *sysmem)
1150 {
1151 DeviceState *dev = DEVICE(flash);
1152
1153 assert(QEMU_IS_ALIGNED(size, VIRT_FLASH_SECTOR_SIZE));
1154 assert(size / VIRT_FLASH_SECTOR_SIZE <= UINT32_MAX);
1155 qdev_prop_set_uint32(dev, "num-blocks", size / VIRT_FLASH_SECTOR_SIZE);
1156 sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1157
1158 memory_region_add_subregion(sysmem, base,
1159 sysbus_mmio_get_region(SYS_BUS_DEVICE(dev),
1160 0));
1161 }
1162
1163 static void virt_flash_map(VirtMachineState *vms,
1164 MemoryRegion *sysmem,
1165 MemoryRegion *secure_sysmem)
1166 {
1167 /*
1168 * Map two flash devices to fill the VIRT_FLASH space in the memmap.
1169 * sysmem is the system memory space. secure_sysmem is the secure view
1170 * of the system, and the first flash device should be made visible only
1171 * there. The second flash device is visible to both secure and nonsecure.
1172 * If sysmem == secure_sysmem this means there is no separate Secure
1173 * address space and both flash devices are generally visible.
1174 */
1175 hwaddr flashsize = vms->memmap[VIRT_FLASH].size / 2;
1176 hwaddr flashbase = vms->memmap[VIRT_FLASH].base;
1177
1178 virt_flash_map1(vms->flash[0], flashbase, flashsize,
1179 secure_sysmem);
1180 virt_flash_map1(vms->flash[1], flashbase + flashsize, flashsize,
1181 sysmem);
1182 }
1183
1184 static void virt_flash_fdt(VirtMachineState *vms,
1185 MemoryRegion *sysmem,
1186 MemoryRegion *secure_sysmem)
1187 {
1188 hwaddr flashsize = vms->memmap[VIRT_FLASH].size / 2;
1189 hwaddr flashbase = vms->memmap[VIRT_FLASH].base;
1190 MachineState *ms = MACHINE(vms);
1191 char *nodename;
1192
1193 if (sysmem == secure_sysmem) {
1194 /* Report both flash devices as a single node in the DT */
1195 nodename = g_strdup_printf("/flash@%" PRIx64, flashbase);
1196 qemu_fdt_add_subnode(ms->fdt, nodename);
1197 qemu_fdt_setprop_string(ms->fdt, nodename, "compatible", "cfi-flash");
1198 qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1199 2, flashbase, 2, flashsize,
1200 2, flashbase + flashsize, 2, flashsize);
1201 qemu_fdt_setprop_cell(ms->fdt, nodename, "bank-width", 4);
1202 g_free(nodename);
1203 } else {
1204 /*
1205 * Report the devices as separate nodes so we can mark one as
1206 * only visible to the secure world.
1207 */
1208 nodename = g_strdup_printf("/secflash@%" PRIx64, flashbase);
1209 qemu_fdt_add_subnode(ms->fdt, nodename);
1210 qemu_fdt_setprop_string(ms->fdt, nodename, "compatible", "cfi-flash");
1211 qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1212 2, flashbase, 2, flashsize);
1213 qemu_fdt_setprop_cell(ms->fdt, nodename, "bank-width", 4);
1214 qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
1215 qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
1216 g_free(nodename);
1217
1218 nodename = g_strdup_printf("/flash@%" PRIx64, flashbase + flashsize);
1219 qemu_fdt_add_subnode(ms->fdt, nodename);
1220 qemu_fdt_setprop_string(ms->fdt, nodename, "compatible", "cfi-flash");
1221 qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1222 2, flashbase + flashsize, 2, flashsize);
1223 qemu_fdt_setprop_cell(ms->fdt, nodename, "bank-width", 4);
1224 g_free(nodename);
1225 }
1226 }
1227
1228 static bool virt_firmware_init(VirtMachineState *vms,
1229 MemoryRegion *sysmem,
1230 MemoryRegion *secure_sysmem)
1231 {
1232 int i;
1233 const char *bios_name;
1234 BlockBackend *pflash_blk0;
1235
1236 /* Map legacy -drive if=pflash to machine properties */
1237 for (i = 0; i < ARRAY_SIZE(vms->flash); i++) {
1238 pflash_cfi01_legacy_drive(vms->flash[i],
1239 drive_get(IF_PFLASH, 0, i));
1240 }
1241
1242 virt_flash_map(vms, sysmem, secure_sysmem);
1243
1244 pflash_blk0 = pflash_cfi01_get_blk(vms->flash[0]);
1245
1246 bios_name = MACHINE(vms)->firmware;
1247 if (bios_name) {
1248 char *fname;
1249 MemoryRegion *mr;
1250 int image_size;
1251
1252 if (pflash_blk0) {
1253 error_report("The contents of the first flash device may be "
1254 "specified with -bios or with -drive if=pflash... "
1255 "but you cannot use both options at once");
1256 exit(1);
1257 }
1258
1259 /* Fall back to -bios */
1260
1261 fname = qemu_find_file(QEMU_FILE_TYPE_BIOS, bios_name);
1262 if (!fname) {
1263 error_report("Could not find ROM image '%s'", bios_name);
1264 exit(1);
1265 }
1266 mr = sysbus_mmio_get_region(SYS_BUS_DEVICE(vms->flash[0]), 0);
1267 image_size = load_image_mr(fname, mr);
1268 g_free(fname);
1269 if (image_size < 0) {
1270 error_report("Could not load ROM image '%s'", bios_name);
1271 exit(1);
1272 }
1273 }
1274
1275 return pflash_blk0 || bios_name;
1276 }
1277
1278 static FWCfgState *create_fw_cfg(const VirtMachineState *vms, AddressSpace *as)
1279 {
1280 MachineState *ms = MACHINE(vms);
1281 hwaddr base = vms->memmap[VIRT_FW_CFG].base;
1282 hwaddr size = vms->memmap[VIRT_FW_CFG].size;
1283 FWCfgState *fw_cfg;
1284 char *nodename;
1285
1286 fw_cfg = fw_cfg_init_mem_wide(base + 8, base, 8, base + 16, as);
1287 fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, (uint16_t)ms->smp.cpus);
1288
1289 nodename = g_strdup_printf("/fw-cfg@%" PRIx64, base);
1290 qemu_fdt_add_subnode(ms->fdt, nodename);
1291 qemu_fdt_setprop_string(ms->fdt, nodename,
1292 "compatible", "qemu,fw-cfg-mmio");
1293 qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1294 2, base, 2, size);
1295 qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1296 g_free(nodename);
1297 return fw_cfg;
1298 }
1299
1300 static void create_pcie_irq_map(const MachineState *ms,
1301 uint32_t gic_phandle,
1302 int first_irq, const char *nodename)
1303 {
1304 int devfn, pin;
1305 uint32_t full_irq_map[4 * 4 * 10] = { 0 };
1306 uint32_t *irq_map = full_irq_map;
1307
1308 for (devfn = 0; devfn <= 0x18; devfn += 0x8) {
1309 for (pin = 0; pin < 4; pin++) {
1310 int irq_type = GIC_FDT_IRQ_TYPE_SPI;
1311 int irq_nr = first_irq + ((pin + PCI_SLOT(devfn)) % PCI_NUM_PINS);
1312 int irq_level = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
1313 int i;
1314
1315 uint32_t map[] = {
1316 devfn << 8, 0, 0, /* devfn */
1317 pin + 1, /* PCI pin */
1318 gic_phandle, 0, 0, irq_type, irq_nr, irq_level }; /* GIC irq */
1319
1320 /* Convert map to big endian */
1321 for (i = 0; i < 10; i++) {
1322 irq_map[i] = cpu_to_be32(map[i]);
1323 }
1324 irq_map += 10;
1325 }
1326 }
1327
1328 qemu_fdt_setprop(ms->fdt, nodename, "interrupt-map",
1329 full_irq_map, sizeof(full_irq_map));
1330
1331 qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupt-map-mask",
1332 cpu_to_be16(PCI_DEVFN(3, 0)), /* Slot 3 */
1333 0, 0,
1334 0x7 /* PCI irq */);
1335 }
1336
1337 static void create_smmu(const VirtMachineState *vms,
1338 PCIBus *bus)
1339 {
1340 char *node;
1341 const char compat[] = "arm,smmu-v3";
1342 int irq = vms->irqmap[VIRT_SMMU];
1343 int i;
1344 hwaddr base = vms->memmap[VIRT_SMMU].base;
1345 hwaddr size = vms->memmap[VIRT_SMMU].size;
1346 const char irq_names[] = "eventq\0priq\0cmdq-sync\0gerror";
1347 DeviceState *dev;
1348 MachineState *ms = MACHINE(vms);
1349
1350 if (vms->iommu != VIRT_IOMMU_SMMUV3 || !vms->iommu_phandle) {
1351 return;
1352 }
1353
1354 dev = qdev_new(TYPE_ARM_SMMUV3);
1355
1356 object_property_set_link(OBJECT(dev), "primary-bus", OBJECT(bus),
1357 &error_abort);
1358 sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1359 sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, base);
1360 for (i = 0; i < NUM_SMMU_IRQS; i++) {
1361 sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
1362 qdev_get_gpio_in(vms->gic, irq + i));
1363 }
1364
1365 node = g_strdup_printf("/smmuv3@%" PRIx64, base);
1366 qemu_fdt_add_subnode(ms->fdt, node);
1367 qemu_fdt_setprop(ms->fdt, node, "compatible", compat, sizeof(compat));
1368 qemu_fdt_setprop_sized_cells(ms->fdt, node, "reg", 2, base, 2, size);
1369
1370 qemu_fdt_setprop_cells(ms->fdt, node, "interrupts",
1371 GIC_FDT_IRQ_TYPE_SPI, irq , GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1372 GIC_FDT_IRQ_TYPE_SPI, irq + 1, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1373 GIC_FDT_IRQ_TYPE_SPI, irq + 2, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1374 GIC_FDT_IRQ_TYPE_SPI, irq + 3, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
1375
1376 qemu_fdt_setprop(ms->fdt, node, "interrupt-names", irq_names,
1377 sizeof(irq_names));
1378
1379 qemu_fdt_setprop(ms->fdt, node, "dma-coherent", NULL, 0);
1380
1381 qemu_fdt_setprop_cell(ms->fdt, node, "#iommu-cells", 1);
1382
1383 qemu_fdt_setprop_cell(ms->fdt, node, "phandle", vms->iommu_phandle);
1384 g_free(node);
1385 }
1386
1387 static void create_virtio_iommu_dt_bindings(VirtMachineState *vms)
1388 {
1389 const char compat[] = "virtio,pci-iommu\0pci1af4,1057";
1390 uint16_t bdf = vms->virtio_iommu_bdf;
1391 MachineState *ms = MACHINE(vms);
1392 char *node;
1393
1394 vms->iommu_phandle = qemu_fdt_alloc_phandle(ms->fdt);
1395
1396 node = g_strdup_printf("%s/virtio_iommu@%x,%x", vms->pciehb_nodename,
1397 PCI_SLOT(bdf), PCI_FUNC(bdf));
1398 qemu_fdt_add_subnode(ms->fdt, node);
1399 qemu_fdt_setprop(ms->fdt, node, "compatible", compat, sizeof(compat));
1400 qemu_fdt_setprop_sized_cells(ms->fdt, node, "reg",
1401 1, bdf << 8, 1, 0, 1, 0,
1402 1, 0, 1, 0);
1403
1404 qemu_fdt_setprop_cell(ms->fdt, node, "#iommu-cells", 1);
1405 qemu_fdt_setprop_cell(ms->fdt, node, "phandle", vms->iommu_phandle);
1406 g_free(node);
1407
1408 qemu_fdt_setprop_cells(ms->fdt, vms->pciehb_nodename, "iommu-map",
1409 0x0, vms->iommu_phandle, 0x0, bdf,
1410 bdf + 1, vms->iommu_phandle, bdf + 1, 0xffff - bdf);
1411 }
1412
1413 static void create_pcie(VirtMachineState *vms)
1414 {
1415 hwaddr base_mmio = vms->memmap[VIRT_PCIE_MMIO].base;
1416 hwaddr size_mmio = vms->memmap[VIRT_PCIE_MMIO].size;
1417 hwaddr base_mmio_high = vms->memmap[VIRT_HIGH_PCIE_MMIO].base;
1418 hwaddr size_mmio_high = vms->memmap[VIRT_HIGH_PCIE_MMIO].size;
1419 hwaddr base_pio = vms->memmap[VIRT_PCIE_PIO].base;
1420 hwaddr size_pio = vms->memmap[VIRT_PCIE_PIO].size;
1421 hwaddr base_ecam, size_ecam;
1422 hwaddr base = base_mmio;
1423 int nr_pcie_buses;
1424 int irq = vms->irqmap[VIRT_PCIE];
1425 MemoryRegion *mmio_alias;
1426 MemoryRegion *mmio_reg;
1427 MemoryRegion *ecam_alias;
1428 MemoryRegion *ecam_reg;
1429 DeviceState *dev;
1430 char *nodename;
1431 int i, ecam_id;
1432 PCIHostState *pci;
1433 MachineState *ms = MACHINE(vms);
1434 MachineClass *mc = MACHINE_GET_CLASS(ms);
1435
1436 dev = qdev_new(TYPE_GPEX_HOST);
1437 sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1438
1439 ecam_id = VIRT_ECAM_ID(vms->highmem_ecam);
1440 base_ecam = vms->memmap[ecam_id].base;
1441 size_ecam = vms->memmap[ecam_id].size;
1442 nr_pcie_buses = size_ecam / PCIE_MMCFG_SIZE_MIN;
1443 /* Map only the first size_ecam bytes of ECAM space */
1444 ecam_alias = g_new0(MemoryRegion, 1);
1445 ecam_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 0);
1446 memory_region_init_alias(ecam_alias, OBJECT(dev), "pcie-ecam",
1447 ecam_reg, 0, size_ecam);
1448 memory_region_add_subregion(get_system_memory(), base_ecam, ecam_alias);
1449
1450 /* Map the MMIO window into system address space so as to expose
1451 * the section of PCI MMIO space which starts at the same base address
1452 * (ie 1:1 mapping for that part of PCI MMIO space visible through
1453 * the window).
1454 */
1455 mmio_alias = g_new0(MemoryRegion, 1);
1456 mmio_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 1);
1457 memory_region_init_alias(mmio_alias, OBJECT(dev), "pcie-mmio",
1458 mmio_reg, base_mmio, size_mmio);
1459 memory_region_add_subregion(get_system_memory(), base_mmio, mmio_alias);
1460
1461 if (vms->highmem_mmio) {
1462 /* Map high MMIO space */
1463 MemoryRegion *high_mmio_alias = g_new0(MemoryRegion, 1);
1464
1465 memory_region_init_alias(high_mmio_alias, OBJECT(dev), "pcie-mmio-high",
1466 mmio_reg, base_mmio_high, size_mmio_high);
1467 memory_region_add_subregion(get_system_memory(), base_mmio_high,
1468 high_mmio_alias);
1469 }
1470
1471 /* Map IO port space */
1472 sysbus_mmio_map(SYS_BUS_DEVICE(dev), 2, base_pio);
1473
1474 for (i = 0; i < GPEX_NUM_IRQS; i++) {
1475 sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
1476 qdev_get_gpio_in(vms->gic, irq + i));
1477 gpex_set_irq_num(GPEX_HOST(dev), i, irq + i);
1478 }
1479
1480 pci = PCI_HOST_BRIDGE(dev);
1481 pci->bypass_iommu = vms->default_bus_bypass_iommu;
1482 vms->bus = pci->bus;
1483 if (vms->bus) {
1484 for (i = 0; i < nb_nics; i++) {
1485 pci_nic_init_nofail(&nd_table[i], pci->bus, mc->default_nic, NULL);
1486 }
1487 }
1488
1489 nodename = vms->pciehb_nodename = g_strdup_printf("/pcie@%" PRIx64, base);
1490 qemu_fdt_add_subnode(ms->fdt, nodename);
1491 qemu_fdt_setprop_string(ms->fdt, nodename,
1492 "compatible", "pci-host-ecam-generic");
1493 qemu_fdt_setprop_string(ms->fdt, nodename, "device_type", "pci");
1494 qemu_fdt_setprop_cell(ms->fdt, nodename, "#address-cells", 3);
1495 qemu_fdt_setprop_cell(ms->fdt, nodename, "#size-cells", 2);
1496 qemu_fdt_setprop_cell(ms->fdt, nodename, "linux,pci-domain", 0);
1497 qemu_fdt_setprop_cells(ms->fdt, nodename, "bus-range", 0,
1498 nr_pcie_buses - 1);
1499 qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1500
1501 if (vms->msi_phandle) {
1502 qemu_fdt_setprop_cells(ms->fdt, nodename, "msi-map",
1503 0, vms->msi_phandle, 0, 0x10000);
1504 }
1505
1506 qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1507 2, base_ecam, 2, size_ecam);
1508
1509 if (vms->highmem_mmio) {
1510 qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "ranges",
1511 1, FDT_PCI_RANGE_IOPORT, 2, 0,
1512 2, base_pio, 2, size_pio,
1513 1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1514 2, base_mmio, 2, size_mmio,
1515 1, FDT_PCI_RANGE_MMIO_64BIT,
1516 2, base_mmio_high,
1517 2, base_mmio_high, 2, size_mmio_high);
1518 } else {
1519 qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "ranges",
1520 1, FDT_PCI_RANGE_IOPORT, 2, 0,
1521 2, base_pio, 2, size_pio,
1522 1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1523 2, base_mmio, 2, size_mmio);
1524 }
1525
1526 qemu_fdt_setprop_cell(ms->fdt, nodename, "#interrupt-cells", 1);
1527 create_pcie_irq_map(ms, vms->gic_phandle, irq, nodename);
1528
1529 if (vms->iommu) {
1530 vms->iommu_phandle = qemu_fdt_alloc_phandle(ms->fdt);
1531
1532 switch (vms->iommu) {
1533 case VIRT_IOMMU_SMMUV3:
1534 create_smmu(vms, vms->bus);
1535 qemu_fdt_setprop_cells(ms->fdt, nodename, "iommu-map",
1536 0x0, vms->iommu_phandle, 0x0, 0x10000);
1537 break;
1538 default:
1539 g_assert_not_reached();
1540 }
1541 }
1542 }
1543
1544 static void create_platform_bus(VirtMachineState *vms)
1545 {
1546 DeviceState *dev;
1547 SysBusDevice *s;
1548 int i;
1549 MemoryRegion *sysmem = get_system_memory();
1550
1551 dev = qdev_new(TYPE_PLATFORM_BUS_DEVICE);
1552 dev->id = g_strdup(TYPE_PLATFORM_BUS_DEVICE);
1553 qdev_prop_set_uint32(dev, "num_irqs", PLATFORM_BUS_NUM_IRQS);
1554 qdev_prop_set_uint32(dev, "mmio_size", vms->memmap[VIRT_PLATFORM_BUS].size);
1555 sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1556 vms->platform_bus_dev = dev;
1557
1558 s = SYS_BUS_DEVICE(dev);
1559 for (i = 0; i < PLATFORM_BUS_NUM_IRQS; i++) {
1560 int irq = vms->irqmap[VIRT_PLATFORM_BUS] + i;
1561 sysbus_connect_irq(s, i, qdev_get_gpio_in(vms->gic, irq));
1562 }
1563
1564 memory_region_add_subregion(sysmem,
1565 vms->memmap[VIRT_PLATFORM_BUS].base,
1566 sysbus_mmio_get_region(s, 0));
1567 }
1568
1569 static void create_tag_ram(MemoryRegion *tag_sysmem,
1570 hwaddr base, hwaddr size,
1571 const char *name)
1572 {
1573 MemoryRegion *tagram = g_new(MemoryRegion, 1);
1574
1575 memory_region_init_ram(tagram, NULL, name, size / 32, &error_fatal);
1576 memory_region_add_subregion(tag_sysmem, base / 32, tagram);
1577 }
1578
1579 static void create_secure_ram(VirtMachineState *vms,
1580 MemoryRegion *secure_sysmem,
1581 MemoryRegion *secure_tag_sysmem)
1582 {
1583 MemoryRegion *secram = g_new(MemoryRegion, 1);
1584 char *nodename;
1585 hwaddr base = vms->memmap[VIRT_SECURE_MEM].base;
1586 hwaddr size = vms->memmap[VIRT_SECURE_MEM].size;
1587 MachineState *ms = MACHINE(vms);
1588
1589 memory_region_init_ram(secram, NULL, "virt.secure-ram", size,
1590 &error_fatal);
1591 memory_region_add_subregion(secure_sysmem, base, secram);
1592
1593 nodename = g_strdup_printf("/secram@%" PRIx64, base);
1594 qemu_fdt_add_subnode(ms->fdt, nodename);
1595 qemu_fdt_setprop_string(ms->fdt, nodename, "device_type", "memory");
1596 qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg", 2, base, 2, size);
1597 qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
1598 qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
1599
1600 if (secure_tag_sysmem) {
1601 create_tag_ram(secure_tag_sysmem, base, size, "mach-virt.secure-tag");
1602 }
1603
1604 g_free(nodename);
1605 }
1606
1607 static void *machvirt_dtb(const struct arm_boot_info *binfo, int *fdt_size)
1608 {
1609 const VirtMachineState *board = container_of(binfo, VirtMachineState,
1610 bootinfo);
1611 MachineState *ms = MACHINE(board);
1612
1613
1614 *fdt_size = board->fdt_size;
1615 return ms->fdt;
1616 }
1617
1618 static void virt_build_smbios(VirtMachineState *vms)
1619 {
1620 MachineClass *mc = MACHINE_GET_CLASS(vms);
1621 MachineState *ms = MACHINE(vms);
1622 VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1623 uint8_t *smbios_tables, *smbios_anchor;
1624 size_t smbios_tables_len, smbios_anchor_len;
1625 struct smbios_phys_mem_area mem_array;
1626 const char *product = "QEMU Virtual Machine";
1627
1628 if (kvm_enabled()) {
1629 product = "KVM Virtual Machine";
1630 }
1631
1632 smbios_set_defaults("QEMU", product,
1633 vmc->smbios_old_sys_ver ? "1.0" : mc->name, false,
1634 true, SMBIOS_ENTRY_POINT_TYPE_64);
1635
1636 /* build the array of physical mem area from base_memmap */
1637 mem_array.address = vms->memmap[VIRT_MEM].base;
1638 mem_array.length = ms->ram_size;
1639
1640 smbios_get_tables(ms, &mem_array, 1,
1641 &smbios_tables, &smbios_tables_len,
1642 &smbios_anchor, &smbios_anchor_len,
1643 &error_fatal);
1644
1645 if (smbios_anchor) {
1646 fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-tables",
1647 smbios_tables, smbios_tables_len);
1648 fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-anchor",
1649 smbios_anchor, smbios_anchor_len);
1650 }
1651 }
1652
1653 static
1654 void virt_machine_done(Notifier *notifier, void *data)
1655 {
1656 VirtMachineState *vms = container_of(notifier, VirtMachineState,
1657 machine_done);
1658 MachineState *ms = MACHINE(vms);
1659 ARMCPU *cpu = ARM_CPU(first_cpu);
1660 struct arm_boot_info *info = &vms->bootinfo;
1661 AddressSpace *as = arm_boot_address_space(cpu, info);
1662
1663 /*
1664 * If the user provided a dtb, we assume the dynamic sysbus nodes
1665 * already are integrated there. This corresponds to a use case where
1666 * the dynamic sysbus nodes are complex and their generation is not yet
1667 * supported. In that case the user can take charge of the guest dt
1668 * while qemu takes charge of the qom stuff.
1669 */
1670 if (info->dtb_filename == NULL) {
1671 platform_bus_add_all_fdt_nodes(ms->fdt, "/intc",
1672 vms->memmap[VIRT_PLATFORM_BUS].base,
1673 vms->memmap[VIRT_PLATFORM_BUS].size,
1674 vms->irqmap[VIRT_PLATFORM_BUS]);
1675 }
1676 if (arm_load_dtb(info->dtb_start, info, info->dtb_limit, as, ms) < 0) {
1677 exit(1);
1678 }
1679
1680 fw_cfg_add_extra_pci_roots(vms->bus, vms->fw_cfg);
1681
1682 virt_acpi_setup(vms);
1683 virt_build_smbios(vms);
1684 }
1685
1686 static uint64_t virt_cpu_mp_affinity(VirtMachineState *vms, int idx)
1687 {
1688 uint8_t clustersz = ARM_DEFAULT_CPUS_PER_CLUSTER;
1689 VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1690
1691 if (!vmc->disallow_affinity_adjustment) {
1692 /* Adjust MPIDR like 64-bit KVM hosts, which incorporate the
1693 * GIC's target-list limitations. 32-bit KVM hosts currently
1694 * always create clusters of 4 CPUs, but that is expected to
1695 * change when they gain support for gicv3. When KVM is enabled
1696 * it will override the changes we make here, therefore our
1697 * purposes are to make TCG consistent (with 64-bit KVM hosts)
1698 * and to improve SGI efficiency.
1699 */
1700 if (vms->gic_version == VIRT_GIC_VERSION_2) {
1701 clustersz = GIC_TARGETLIST_BITS;
1702 } else {
1703 clustersz = GICV3_TARGETLIST_BITS;
1704 }
1705 }
1706 return arm_cpu_mp_affinity(idx, clustersz);
1707 }
1708
1709 static inline bool *virt_get_high_memmap_enabled(VirtMachineState *vms,
1710 int index)
1711 {
1712 bool *enabled_array[] = {
1713 &vms->highmem_redists,
1714 &vms->highmem_ecam,
1715 &vms->highmem_mmio,
1716 };
1717
1718 assert(ARRAY_SIZE(extended_memmap) - VIRT_LOWMEMMAP_LAST ==
1719 ARRAY_SIZE(enabled_array));
1720 assert(index - VIRT_LOWMEMMAP_LAST < ARRAY_SIZE(enabled_array));
1721
1722 return enabled_array[index - VIRT_LOWMEMMAP_LAST];
1723 }
1724
1725 static void virt_set_high_memmap(VirtMachineState *vms,
1726 hwaddr base, int pa_bits)
1727 {
1728 hwaddr region_base, region_size;
1729 bool *region_enabled, fits;
1730 int i;
1731
1732 for (i = VIRT_LOWMEMMAP_LAST; i < ARRAY_SIZE(extended_memmap); i++) {
1733 region_enabled = virt_get_high_memmap_enabled(vms, i);
1734 region_base = ROUND_UP(base, extended_memmap[i].size);
1735 region_size = extended_memmap[i].size;
1736
1737 vms->memmap[i].base = region_base;
1738 vms->memmap[i].size = region_size;
1739
1740 /*
1741 * Check each device to see if it fits in the PA space,
1742 * moving highest_gpa as we go. For compatibility, move
1743 * highest_gpa for disabled fitting devices as well, if
1744 * the compact layout has been disabled.
1745 *
1746 * For each device that doesn't fit, disable it.
1747 */
1748 fits = (region_base + region_size) <= BIT_ULL(pa_bits);
1749 *region_enabled &= fits;
1750 if (vms->highmem_compact && !*region_enabled) {
1751 continue;
1752 }
1753
1754 base = region_base + region_size;
1755 if (fits) {
1756 vms->highest_gpa = base - 1;
1757 }
1758 }
1759 }
1760
1761 static void virt_set_memmap(VirtMachineState *vms, int pa_bits)
1762 {
1763 MachineState *ms = MACHINE(vms);
1764 hwaddr base, device_memory_base, device_memory_size, memtop;
1765 int i;
1766
1767 vms->memmap = extended_memmap;
1768
1769 for (i = 0; i < ARRAY_SIZE(base_memmap); i++) {
1770 vms->memmap[i] = base_memmap[i];
1771 }
1772
1773 if (ms->ram_slots > ACPI_MAX_RAM_SLOTS) {
1774 error_report("unsupported number of memory slots: %"PRIu64,
1775 ms->ram_slots);
1776 exit(EXIT_FAILURE);
1777 }
1778
1779 /*
1780 * !highmem is exactly the same as limiting the PA space to 32bit,
1781 * irrespective of the underlying capabilities of the HW.
1782 */
1783 if (!vms->highmem) {
1784 pa_bits = 32;
1785 }
1786
1787 /*
1788 * We compute the base of the high IO region depending on the
1789 * amount of initial and device memory. The device memory start/size
1790 * is aligned on 1GiB. We never put the high IO region below 256GiB
1791 * so that if maxram_size is < 255GiB we keep the legacy memory map.
1792 * The device region size assumes 1GiB page max alignment per slot.
1793 */
1794 device_memory_base =
1795 ROUND_UP(vms->memmap[VIRT_MEM].base + ms->ram_size, GiB);
1796 device_memory_size = ms->maxram_size - ms->ram_size + ms->ram_slots * GiB;
1797
1798 /* Base address of the high IO region */
1799 memtop = base = device_memory_base + ROUND_UP(device_memory_size, GiB);
1800 if (memtop > BIT_ULL(pa_bits)) {
1801 error_report("Addressing limited to %d bits, but memory exceeds it by %llu bytes\n",
1802 pa_bits, memtop - BIT_ULL(pa_bits));
1803 exit(EXIT_FAILURE);
1804 }
1805 if (base < device_memory_base) {
1806 error_report("maxmem/slots too huge");
1807 exit(EXIT_FAILURE);
1808 }
1809 if (base < vms->memmap[VIRT_MEM].base + LEGACY_RAMLIMIT_BYTES) {
1810 base = vms->memmap[VIRT_MEM].base + LEGACY_RAMLIMIT_BYTES;
1811 }
1812
1813 /* We know for sure that at least the memory fits in the PA space */
1814 vms->highest_gpa = memtop - 1;
1815
1816 virt_set_high_memmap(vms, base, pa_bits);
1817
1818 if (device_memory_size > 0) {
1819 machine_memory_devices_init(ms, device_memory_base, device_memory_size);
1820 }
1821 }
1822
1823 static VirtGICType finalize_gic_version_do(const char *accel_name,
1824 VirtGICType gic_version,
1825 int gics_supported,
1826 unsigned int max_cpus)
1827 {
1828 /* Convert host/max/nosel to GIC version number */
1829 switch (gic_version) {
1830 case VIRT_GIC_VERSION_HOST:
1831 if (!kvm_enabled()) {
1832 error_report("gic-version=host requires KVM");
1833 exit(1);
1834 }
1835
1836 /* For KVM, gic-version=host means gic-version=max */
1837 return finalize_gic_version_do(accel_name, VIRT_GIC_VERSION_MAX,
1838 gics_supported, max_cpus);
1839 case VIRT_GIC_VERSION_MAX:
1840 if (gics_supported & VIRT_GIC_VERSION_4_MASK) {
1841 gic_version = VIRT_GIC_VERSION_4;
1842 } else if (gics_supported & VIRT_GIC_VERSION_3_MASK) {
1843 gic_version = VIRT_GIC_VERSION_3;
1844 } else {
1845 gic_version = VIRT_GIC_VERSION_2;
1846 }
1847 break;
1848 case VIRT_GIC_VERSION_NOSEL:
1849 if ((gics_supported & VIRT_GIC_VERSION_2_MASK) &&
1850 max_cpus <= GIC_NCPU) {
1851 gic_version = VIRT_GIC_VERSION_2;
1852 } else if (gics_supported & VIRT_GIC_VERSION_3_MASK) {
1853 /*
1854 * in case the host does not support v2 emulation or
1855 * the end-user requested more than 8 VCPUs we now default
1856 * to v3. In any case defaulting to v2 would be broken.
1857 */
1858 gic_version = VIRT_GIC_VERSION_3;
1859 } else if (max_cpus > GIC_NCPU) {
1860 error_report("%s only supports GICv2 emulation but more than 8 "
1861 "vcpus are requested", accel_name);
1862 exit(1);
1863 }
1864 break;
1865 case VIRT_GIC_VERSION_2:
1866 case VIRT_GIC_VERSION_3:
1867 case VIRT_GIC_VERSION_4:
1868 break;
1869 }
1870
1871 /* Check chosen version is effectively supported */
1872 switch (gic_version) {
1873 case VIRT_GIC_VERSION_2:
1874 if (!(gics_supported & VIRT_GIC_VERSION_2_MASK)) {
1875 error_report("%s does not support GICv2 emulation", accel_name);
1876 exit(1);
1877 }
1878 break;
1879 case VIRT_GIC_VERSION_3:
1880 if (!(gics_supported & VIRT_GIC_VERSION_3_MASK)) {
1881 error_report("%s does not support GICv3 emulation", accel_name);
1882 exit(1);
1883 }
1884 break;
1885 case VIRT_GIC_VERSION_4:
1886 if (!(gics_supported & VIRT_GIC_VERSION_4_MASK)) {
1887 error_report("%s does not support GICv4 emulation, is virtualization=on?",
1888 accel_name);
1889 exit(1);
1890 }
1891 break;
1892 default:
1893 error_report("logic error in finalize_gic_version");
1894 exit(1);
1895 break;
1896 }
1897
1898 return gic_version;
1899 }
1900
1901 /*
1902 * finalize_gic_version - Determines the final gic_version
1903 * according to the gic-version property
1904 *
1905 * Default GIC type is v2
1906 */
1907 static void finalize_gic_version(VirtMachineState *vms)
1908 {
1909 const char *accel_name = current_accel_name();
1910 unsigned int max_cpus = MACHINE(vms)->smp.max_cpus;
1911 int gics_supported = 0;
1912
1913 /* Determine which GIC versions the current environment supports */
1914 if (kvm_enabled() && kvm_irqchip_in_kernel()) {
1915 int probe_bitmap = kvm_arm_vgic_probe();
1916
1917 if (!probe_bitmap) {
1918 error_report("Unable to determine GIC version supported by host");
1919 exit(1);
1920 }
1921
1922 if (probe_bitmap & KVM_ARM_VGIC_V2) {
1923 gics_supported |= VIRT_GIC_VERSION_2_MASK;
1924 }
1925 if (probe_bitmap & KVM_ARM_VGIC_V3) {
1926 gics_supported |= VIRT_GIC_VERSION_3_MASK;
1927 }
1928 } else if (kvm_enabled() && !kvm_irqchip_in_kernel()) {
1929 /* KVM w/o kernel irqchip can only deal with GICv2 */
1930 gics_supported |= VIRT_GIC_VERSION_2_MASK;
1931 accel_name = "KVM with kernel-irqchip=off";
1932 } else if (tcg_enabled() || hvf_enabled() || qtest_enabled()) {
1933 gics_supported |= VIRT_GIC_VERSION_2_MASK;
1934 if (module_object_class_by_name("arm-gicv3")) {
1935 gics_supported |= VIRT_GIC_VERSION_3_MASK;
1936 if (vms->virt) {
1937 /* GICv4 only makes sense if CPU has EL2 */
1938 gics_supported |= VIRT_GIC_VERSION_4_MASK;
1939 }
1940 }
1941 } else {
1942 error_report("Unsupported accelerator, can not determine GIC support");
1943 exit(1);
1944 }
1945
1946 /*
1947 * Then convert helpers like host/max to concrete GIC versions and ensure
1948 * the desired version is supported
1949 */
1950 vms->gic_version = finalize_gic_version_do(accel_name, vms->gic_version,
1951 gics_supported, max_cpus);
1952 }
1953
1954 /*
1955 * virt_cpu_post_init() must be called after the CPUs have
1956 * been realized and the GIC has been created.
1957 */
1958 static void virt_cpu_post_init(VirtMachineState *vms, MemoryRegion *sysmem)
1959 {
1960 int max_cpus = MACHINE(vms)->smp.max_cpus;
1961 bool aarch64, pmu, steal_time;
1962 CPUState *cpu;
1963
1964 aarch64 = object_property_get_bool(OBJECT(first_cpu), "aarch64", NULL);
1965 pmu = object_property_get_bool(OBJECT(first_cpu), "pmu", NULL);
1966 steal_time = object_property_get_bool(OBJECT(first_cpu),
1967 "kvm-steal-time", NULL);
1968
1969 if (kvm_enabled()) {
1970 hwaddr pvtime_reg_base = vms->memmap[VIRT_PVTIME].base;
1971 hwaddr pvtime_reg_size = vms->memmap[VIRT_PVTIME].size;
1972
1973 if (steal_time) {
1974 MemoryRegion *pvtime = g_new(MemoryRegion, 1);
1975 hwaddr pvtime_size = max_cpus * PVTIME_SIZE_PER_CPU;
1976
1977 /* The memory region size must be a multiple of host page size. */
1978 pvtime_size = REAL_HOST_PAGE_ALIGN(pvtime_size);
1979
1980 if (pvtime_size > pvtime_reg_size) {
1981 error_report("pvtime requires a %" HWADDR_PRId
1982 " byte memory region for %d CPUs,"
1983 " but only %" HWADDR_PRId " has been reserved",
1984 pvtime_size, max_cpus, pvtime_reg_size);
1985 exit(1);
1986 }
1987
1988 memory_region_init_ram(pvtime, NULL, "pvtime", pvtime_size, NULL);
1989 memory_region_add_subregion(sysmem, pvtime_reg_base, pvtime);
1990 }
1991
1992 CPU_FOREACH(cpu) {
1993 if (pmu) {
1994 assert(arm_feature(&ARM_CPU(cpu)->env, ARM_FEATURE_PMU));
1995 if (kvm_irqchip_in_kernel()) {
1996 kvm_arm_pmu_set_irq(cpu, VIRTUAL_PMU_IRQ);
1997 }
1998 kvm_arm_pmu_init(cpu);
1999 }
2000 if (steal_time) {
2001 kvm_arm_pvtime_init(cpu, pvtime_reg_base +
2002 cpu->cpu_index * PVTIME_SIZE_PER_CPU);
2003 }
2004 }
2005 } else {
2006 if (aarch64 && vms->highmem) {
2007 int requested_pa_size = 64 - clz64(vms->highest_gpa);
2008 int pamax = arm_pamax(ARM_CPU(first_cpu));
2009
2010 if (pamax < requested_pa_size) {
2011 error_report("VCPU supports less PA bits (%d) than "
2012 "requested by the memory map (%d)",
2013 pamax, requested_pa_size);
2014 exit(1);
2015 }
2016 }
2017 }
2018 }
2019
2020 static void machvirt_init(MachineState *machine)
2021 {
2022 VirtMachineState *vms = VIRT_MACHINE(machine);
2023 VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(machine);
2024 MachineClass *mc = MACHINE_GET_CLASS(machine);
2025 const CPUArchIdList *possible_cpus;
2026 MemoryRegion *sysmem = get_system_memory();
2027 MemoryRegion *secure_sysmem = NULL;
2028 MemoryRegion *tag_sysmem = NULL;
2029 MemoryRegion *secure_tag_sysmem = NULL;
2030 int n, virt_max_cpus;
2031 bool firmware_loaded;
2032 bool aarch64 = true;
2033 bool has_ged = !vmc->no_ged;
2034 unsigned int smp_cpus = machine->smp.cpus;
2035 unsigned int max_cpus = machine->smp.max_cpus;
2036
2037 if (!cpu_type_valid(machine->cpu_type)) {
2038 error_report("mach-virt: CPU type %s not supported", machine->cpu_type);
2039 exit(1);
2040 }
2041
2042 possible_cpus = mc->possible_cpu_arch_ids(machine);
2043
2044 /*
2045 * In accelerated mode, the memory map is computed earlier in kvm_type()
2046 * to create a VM with the right number of IPA bits.
2047 */
2048 if (!vms->memmap) {
2049 Object *cpuobj;
2050 ARMCPU *armcpu;
2051 int pa_bits;
2052
2053 /*
2054 * Instantiate a temporary CPU object to find out about what
2055 * we are about to deal with. Once this is done, get rid of
2056 * the object.
2057 */
2058 cpuobj = object_new(possible_cpus->cpus[0].type);
2059 armcpu = ARM_CPU(cpuobj);
2060
2061 pa_bits = arm_pamax(armcpu);
2062
2063 object_unref(cpuobj);
2064
2065 virt_set_memmap(vms, pa_bits);
2066 }
2067
2068 /* We can probe only here because during property set
2069 * KVM is not available yet
2070 */
2071 finalize_gic_version(vms);
2072
2073 if (vms->secure) {
2074 /*
2075 * The Secure view of the world is the same as the NonSecure,
2076 * but with a few extra devices. Create it as a container region
2077 * containing the system memory at low priority; any secure-only
2078 * devices go in at higher priority and take precedence.
2079 */
2080 secure_sysmem = g_new(MemoryRegion, 1);
2081 memory_region_init(secure_sysmem, OBJECT(machine), "secure-memory",
2082 UINT64_MAX);
2083 memory_region_add_subregion_overlap(secure_sysmem, 0, sysmem, -1);
2084 }
2085
2086 firmware_loaded = virt_firmware_init(vms, sysmem,
2087 secure_sysmem ?: sysmem);
2088
2089 /* If we have an EL3 boot ROM then the assumption is that it will
2090 * implement PSCI itself, so disable QEMU's internal implementation
2091 * so it doesn't get in the way. Instead of starting secondary
2092 * CPUs in PSCI powerdown state we will start them all running and
2093 * let the boot ROM sort them out.
2094 * The usual case is that we do use QEMU's PSCI implementation;
2095 * if the guest has EL2 then we will use SMC as the conduit,
2096 * and otherwise we will use HVC (for backwards compatibility and
2097 * because if we're using KVM then we must use HVC).
2098 */
2099 if (vms->secure && firmware_loaded) {
2100 vms->psci_conduit = QEMU_PSCI_CONDUIT_DISABLED;
2101 } else if (vms->virt) {
2102 vms->psci_conduit = QEMU_PSCI_CONDUIT_SMC;
2103 } else {
2104 vms->psci_conduit = QEMU_PSCI_CONDUIT_HVC;
2105 }
2106
2107 /*
2108 * The maximum number of CPUs depends on the GIC version, or on how
2109 * many redistributors we can fit into the memory map (which in turn
2110 * depends on whether this is a GICv3 or v4).
2111 */
2112 if (vms->gic_version == VIRT_GIC_VERSION_2) {
2113 virt_max_cpus = GIC_NCPU;
2114 } else {
2115 virt_max_cpus = virt_redist_capacity(vms, VIRT_GIC_REDIST);
2116 if (vms->highmem_redists) {
2117 virt_max_cpus += virt_redist_capacity(vms, VIRT_HIGH_GIC_REDIST2);
2118 }
2119 }
2120
2121 if (max_cpus > virt_max_cpus) {
2122 error_report("Number of SMP CPUs requested (%d) exceeds max CPUs "
2123 "supported by machine 'mach-virt' (%d)",
2124 max_cpus, virt_max_cpus);
2125 if (vms->gic_version != VIRT_GIC_VERSION_2 && !vms->highmem_redists) {
2126 error_printf("Try 'highmem-redists=on' for more CPUs\n");
2127 }
2128
2129 exit(1);
2130 }
2131
2132 if (vms->secure && (kvm_enabled() || hvf_enabled())) {
2133 error_report("mach-virt: %s does not support providing "
2134 "Security extensions (TrustZone) to the guest CPU",
2135 current_accel_name());
2136 exit(1);
2137 }
2138
2139 if (vms->virt && (kvm_enabled() || hvf_enabled())) {
2140 error_report("mach-virt: %s does not support providing "
2141 "Virtualization extensions to the guest CPU",
2142 current_accel_name());
2143 exit(1);
2144 }
2145
2146 if (vms->mte && (kvm_enabled() || hvf_enabled())) {
2147 error_report("mach-virt: %s does not support providing "
2148 "MTE to the guest CPU",
2149 current_accel_name());
2150 exit(1);
2151 }
2152
2153 create_fdt(vms);
2154
2155 assert(possible_cpus->len == max_cpus);
2156 for (n = 0; n < possible_cpus->len; n++) {
2157 Object *cpuobj;
2158 CPUState *cs;
2159
2160 if (n >= smp_cpus) {
2161 break;
2162 }
2163
2164 cpuobj = object_new(possible_cpus->cpus[n].type);
2165 object_property_set_int(cpuobj, "mp-affinity",
2166 possible_cpus->cpus[n].arch_id, NULL);
2167
2168 cs = CPU(cpuobj);
2169 cs->cpu_index = n;
2170
2171 numa_cpu_pre_plug(&possible_cpus->cpus[cs->cpu_index], DEVICE(cpuobj),
2172 &error_fatal);
2173
2174 aarch64 &= object_property_get_bool(cpuobj, "aarch64", NULL);
2175
2176 if (!vms->secure) {
2177 object_property_set_bool(cpuobj, "has_el3", false, NULL);
2178 }
2179
2180 if (!vms->virt && object_property_find(cpuobj, "has_el2")) {
2181 object_property_set_bool(cpuobj, "has_el2", false, NULL);
2182 }
2183
2184 if (vmc->kvm_no_adjvtime &&
2185 object_property_find(cpuobj, "kvm-no-adjvtime")) {
2186 object_property_set_bool(cpuobj, "kvm-no-adjvtime", true, NULL);
2187 }
2188
2189 if (vmc->no_kvm_steal_time &&
2190 object_property_find(cpuobj, "kvm-steal-time")) {
2191 object_property_set_bool(cpuobj, "kvm-steal-time", false, NULL);
2192 }
2193
2194 if (vmc->no_pmu && object_property_find(cpuobj, "pmu")) {
2195 object_property_set_bool(cpuobj, "pmu", false, NULL);
2196 }
2197
2198 if (vmc->no_tcg_lpa2 && object_property_find(cpuobj, "lpa2")) {
2199 object_property_set_bool(cpuobj, "lpa2", false, NULL);
2200 }
2201
2202 if (object_property_find(cpuobj, "reset-cbar")) {
2203 object_property_set_int(cpuobj, "reset-cbar",
2204 vms->memmap[VIRT_CPUPERIPHS].base,
2205 &error_abort);
2206 }
2207
2208 object_property_set_link(cpuobj, "memory", OBJECT(sysmem),
2209 &error_abort);
2210 if (vms->secure) {
2211 object_property_set_link(cpuobj, "secure-memory",
2212 OBJECT(secure_sysmem), &error_abort);
2213 }
2214
2215 if (vms->mte) {
2216 /* Create the memory region only once, but link to all cpus. */
2217 if (!tag_sysmem) {
2218 /*
2219 * The property exists only if MemTag is supported.
2220 * If it is, we must allocate the ram to back that up.
2221 */
2222 if (!object_property_find(cpuobj, "tag-memory")) {
2223 error_report("MTE requested, but not supported "
2224 "by the guest CPU");
2225 exit(1);
2226 }
2227
2228 tag_sysmem = g_new(MemoryRegion, 1);
2229 memory_region_init(tag_sysmem, OBJECT(machine),
2230 "tag-memory", UINT64_MAX / 32);
2231
2232 if (vms->secure) {
2233 secure_tag_sysmem = g_new(MemoryRegion, 1);
2234 memory_region_init(secure_tag_sysmem, OBJECT(machine),
2235 "secure-tag-memory", UINT64_MAX / 32);
2236
2237 /* As with ram, secure-tag takes precedence over tag. */
2238 memory_region_add_subregion_overlap(secure_tag_sysmem, 0,
2239 tag_sysmem, -1);
2240 }
2241 }
2242
2243 object_property_set_link(cpuobj, "tag-memory", OBJECT(tag_sysmem),
2244 &error_abort);
2245 if (vms->secure) {
2246 object_property_set_link(cpuobj, "secure-tag-memory",
2247 OBJECT(secure_tag_sysmem),
2248 &error_abort);
2249 }
2250 }
2251
2252 qdev_realize(DEVICE(cpuobj), NULL, &error_fatal);
2253 object_unref(cpuobj);
2254 }
2255 fdt_add_timer_nodes(vms);
2256 fdt_add_cpu_nodes(vms);
2257
2258 memory_region_add_subregion(sysmem, vms->memmap[VIRT_MEM].base,
2259 machine->ram);
2260
2261 virt_flash_fdt(vms, sysmem, secure_sysmem ?: sysmem);
2262
2263 create_gic(vms, sysmem);
2264
2265 virt_cpu_post_init(vms, sysmem);
2266
2267 fdt_add_pmu_nodes(vms);
2268
2269 create_uart(vms, VIRT_UART, sysmem, serial_hd(0));
2270
2271 if (vms->secure) {
2272 create_secure_ram(vms, secure_sysmem, secure_tag_sysmem);
2273 create_uart(vms, VIRT_SECURE_UART, secure_sysmem, serial_hd(1));
2274 }
2275
2276 if (tag_sysmem) {
2277 create_tag_ram(tag_sysmem, vms->memmap[VIRT_MEM].base,
2278 machine->ram_size, "mach-virt.tag");
2279 }
2280
2281 vms->highmem_ecam &= (!firmware_loaded || aarch64);
2282
2283 create_rtc(vms);
2284
2285 create_pcie(vms);
2286
2287 if (has_ged && aarch64 && firmware_loaded && virt_is_acpi_enabled(vms)) {
2288 vms->acpi_dev = create_acpi_ged(vms);
2289 } else {
2290 create_gpio_devices(vms, VIRT_GPIO, sysmem);
2291 }
2292
2293 if (vms->secure && !vmc->no_secure_gpio) {
2294 create_gpio_devices(vms, VIRT_SECURE_GPIO, secure_sysmem);
2295 }
2296
2297 /* connect powerdown request */
2298 vms->powerdown_notifier.notify = virt_powerdown_req;
2299 qemu_register_powerdown_notifier(&vms->powerdown_notifier);
2300
2301 /* Create mmio transports, so the user can create virtio backends
2302 * (which will be automatically plugged in to the transports). If
2303 * no backend is created the transport will just sit harmlessly idle.
2304 */
2305 create_virtio_devices(vms);
2306
2307 vms->fw_cfg = create_fw_cfg(vms, &address_space_memory);
2308 rom_set_fw(vms->fw_cfg);
2309
2310 create_platform_bus(vms);
2311
2312 if (machine->nvdimms_state->is_enabled) {
2313 const struct AcpiGenericAddress arm_virt_nvdimm_acpi_dsmio = {
2314 .space_id = AML_AS_SYSTEM_MEMORY,
2315 .address = vms->memmap[VIRT_NVDIMM_ACPI].base,
2316 .bit_width = NVDIMM_ACPI_IO_LEN << 3
2317 };
2318
2319 nvdimm_init_acpi_state(machine->nvdimms_state, sysmem,
2320 arm_virt_nvdimm_acpi_dsmio,
2321 vms->fw_cfg, OBJECT(vms));
2322 }
2323
2324 vms->bootinfo.ram_size = machine->ram_size;
2325 vms->bootinfo.board_id = -1;
2326 vms->bootinfo.loader_start = vms->memmap[VIRT_MEM].base;
2327 vms->bootinfo.get_dtb = machvirt_dtb;
2328 vms->bootinfo.skip_dtb_autoload = true;
2329 vms->bootinfo.firmware_loaded = firmware_loaded;
2330 vms->bootinfo.psci_conduit = vms->psci_conduit;
2331 arm_load_kernel(ARM_CPU(first_cpu), machine, &vms->bootinfo);
2332
2333 vms->machine_done.notify = virt_machine_done;
2334 qemu_add_machine_init_done_notifier(&vms->machine_done);
2335 }
2336
2337 static bool virt_get_secure(Object *obj, Error **errp)
2338 {
2339 VirtMachineState *vms = VIRT_MACHINE(obj);
2340
2341 return vms->secure;
2342 }
2343
2344 static void virt_set_secure(Object *obj, bool value, Error **errp)
2345 {
2346 VirtMachineState *vms = VIRT_MACHINE(obj);
2347
2348 vms->secure = value;
2349 }
2350
2351 static bool virt_get_virt(Object *obj, Error **errp)
2352 {
2353 VirtMachineState *vms = VIRT_MACHINE(obj);
2354
2355 return vms->virt;
2356 }
2357
2358 static void virt_set_virt(Object *obj, bool value, Error **errp)
2359 {
2360 VirtMachineState *vms = VIRT_MACHINE(obj);
2361
2362 vms->virt = value;
2363 }
2364
2365 static bool virt_get_highmem(Object *obj, Error **errp)
2366 {
2367 VirtMachineState *vms = VIRT_MACHINE(obj);
2368
2369 return vms->highmem;
2370 }
2371
2372 static void virt_set_highmem(Object *obj, bool value, Error **errp)
2373 {
2374 VirtMachineState *vms = VIRT_MACHINE(obj);
2375
2376 vms->highmem = value;
2377 }
2378
2379 static bool virt_get_compact_highmem(Object *obj, Error **errp)
2380 {
2381 VirtMachineState *vms = VIRT_MACHINE(obj);
2382
2383 return vms->highmem_compact;
2384 }
2385
2386 static void virt_set_compact_highmem(Object *obj, bool value, Error **errp)
2387 {
2388 VirtMachineState *vms = VIRT_MACHINE(obj);
2389
2390 vms->highmem_compact = value;
2391 }
2392
2393 static bool virt_get_highmem_redists(Object *obj, Error **errp)
2394 {
2395 VirtMachineState *vms = VIRT_MACHINE(obj);
2396
2397 return vms->highmem_redists;
2398 }
2399
2400 static void virt_set_highmem_redists(Object *obj, bool value, Error **errp)
2401 {
2402 VirtMachineState *vms = VIRT_MACHINE(obj);
2403
2404 vms->highmem_redists = value;
2405 }
2406
2407 static bool virt_get_highmem_ecam(Object *obj, Error **errp)
2408 {
2409 VirtMachineState *vms = VIRT_MACHINE(obj);
2410
2411 return vms->highmem_ecam;
2412 }
2413
2414 static void virt_set_highmem_ecam(Object *obj, bool value, Error **errp)
2415 {
2416 VirtMachineState *vms = VIRT_MACHINE(obj);
2417
2418 vms->highmem_ecam = value;
2419 }
2420
2421 static bool virt_get_highmem_mmio(Object *obj, Error **errp)
2422 {
2423 VirtMachineState *vms = VIRT_MACHINE(obj);
2424
2425 return vms->highmem_mmio;
2426 }
2427
2428 static void virt_set_highmem_mmio(Object *obj, bool value, Error **errp)
2429 {
2430 VirtMachineState *vms = VIRT_MACHINE(obj);
2431
2432 vms->highmem_mmio = value;
2433 }
2434
2435
2436 static bool virt_get_its(Object *obj, Error **errp)
2437 {
2438 VirtMachineState *vms = VIRT_MACHINE(obj);
2439
2440 return vms->its;
2441 }
2442
2443 static void virt_set_its(Object *obj, bool value, Error **errp)
2444 {
2445 VirtMachineState *vms = VIRT_MACHINE(obj);
2446
2447 vms->its = value;
2448 }
2449
2450 static bool virt_get_dtb_randomness(Object *obj, Error **errp)
2451 {
2452 VirtMachineState *vms = VIRT_MACHINE(obj);
2453
2454 return vms->dtb_randomness;
2455 }
2456
2457 static void virt_set_dtb_randomness(Object *obj, bool value, Error **errp)
2458 {
2459 VirtMachineState *vms = VIRT_MACHINE(obj);
2460
2461 vms->dtb_randomness = value;
2462 }
2463
2464 static char *virt_get_oem_id(Object *obj, Error **errp)
2465 {
2466 VirtMachineState *vms = VIRT_MACHINE(obj);
2467
2468 return g_strdup(vms->oem_id);
2469 }
2470
2471 static void virt_set_oem_id(Object *obj, const char *value, Error **errp)
2472 {
2473 VirtMachineState *vms = VIRT_MACHINE(obj);
2474 size_t len = strlen(value);
2475
2476 if (len > 6) {
2477 error_setg(errp,
2478 "User specified oem-id value is bigger than 6 bytes in size");
2479 return;
2480 }
2481
2482 strncpy(vms->oem_id, value, 6);
2483 }
2484
2485 static char *virt_get_oem_table_id(Object *obj, Error **errp)
2486 {
2487 VirtMachineState *vms = VIRT_MACHINE(obj);
2488
2489 return g_strdup(vms->oem_table_id);
2490 }
2491
2492 static void virt_set_oem_table_id(Object *obj, const char *value,
2493 Error **errp)
2494 {
2495 VirtMachineState *vms = VIRT_MACHINE(obj);
2496 size_t len = strlen(value);
2497
2498 if (len > 8) {
2499 error_setg(errp,
2500 "User specified oem-table-id value is bigger than 8 bytes in size");
2501 return;
2502 }
2503 strncpy(vms->oem_table_id, value, 8);
2504 }
2505
2506
2507 bool virt_is_acpi_enabled(VirtMachineState *vms)
2508 {
2509 if (vms->acpi == ON_OFF_AUTO_OFF) {
2510 return false;
2511 }
2512 return true;
2513 }
2514
2515 static void virt_get_acpi(Object *obj, Visitor *v, const char *name,
2516 void *opaque, Error **errp)
2517 {
2518 VirtMachineState *vms = VIRT_MACHINE(obj);
2519 OnOffAuto acpi = vms->acpi;
2520
2521 visit_type_OnOffAuto(v, name, &acpi, errp);
2522 }
2523
2524 static void virt_set_acpi(Object *obj, Visitor *v, const char *name,
2525 void *opaque, Error **errp)
2526 {
2527 VirtMachineState *vms = VIRT_MACHINE(obj);
2528
2529 visit_type_OnOffAuto(v, name, &vms->acpi, errp);
2530 }
2531
2532 static bool virt_get_ras(Object *obj, Error **errp)
2533 {
2534 VirtMachineState *vms = VIRT_MACHINE(obj);
2535
2536 return vms->ras;
2537 }
2538
2539 static void virt_set_ras(Object *obj, bool value, Error **errp)
2540 {
2541 VirtMachineState *vms = VIRT_MACHINE(obj);
2542
2543 vms->ras = value;
2544 }
2545
2546 static bool virt_get_mte(Object *obj, Error **errp)
2547 {
2548 VirtMachineState *vms = VIRT_MACHINE(obj);
2549
2550 return vms->mte;
2551 }
2552
2553 static void virt_set_mte(Object *obj, bool value, Error **errp)
2554 {
2555 VirtMachineState *vms = VIRT_MACHINE(obj);
2556
2557 vms->mte = value;
2558 }
2559
2560 static char *virt_get_gic_version(Object *obj, Error **errp)
2561 {
2562 VirtMachineState *vms = VIRT_MACHINE(obj);
2563 const char *val;
2564
2565 switch (vms->gic_version) {
2566 case VIRT_GIC_VERSION_4:
2567 val = "4";
2568 break;
2569 case VIRT_GIC_VERSION_3:
2570 val = "3";
2571 break;
2572 default:
2573 val = "2";
2574 break;
2575 }
2576 return g_strdup(val);
2577 }
2578
2579 static void virt_set_gic_version(Object *obj, const char *value, Error **errp)
2580 {
2581 VirtMachineState *vms = VIRT_MACHINE(obj);
2582
2583 if (!strcmp(value, "4")) {
2584 vms->gic_version = VIRT_GIC_VERSION_4;
2585 } else if (!strcmp(value, "3")) {
2586 vms->gic_version = VIRT_GIC_VERSION_3;
2587 } else if (!strcmp(value, "2")) {
2588 vms->gic_version = VIRT_GIC_VERSION_2;
2589 } else if (!strcmp(value, "host")) {
2590 vms->gic_version = VIRT_GIC_VERSION_HOST; /* Will probe later */
2591 } else if (!strcmp(value, "max")) {
2592 vms->gic_version = VIRT_GIC_VERSION_MAX; /* Will probe later */
2593 } else {
2594 error_setg(errp, "Invalid gic-version value");
2595 error_append_hint(errp, "Valid values are 3, 2, host, max.\n");
2596 }
2597 }
2598
2599 static char *virt_get_iommu(Object *obj, Error **errp)
2600 {
2601 VirtMachineState *vms = VIRT_MACHINE(obj);
2602
2603 switch (vms->iommu) {
2604 case VIRT_IOMMU_NONE:
2605 return g_strdup("none");
2606 case VIRT_IOMMU_SMMUV3:
2607 return g_strdup("smmuv3");
2608 default:
2609 g_assert_not_reached();
2610 }
2611 }
2612
2613 static void virt_set_iommu(Object *obj, const char *value, Error **errp)
2614 {
2615 VirtMachineState *vms = VIRT_MACHINE(obj);
2616
2617 if (!strcmp(value, "smmuv3")) {
2618 vms->iommu = VIRT_IOMMU_SMMUV3;
2619 } else if (!strcmp(value, "none")) {
2620 vms->iommu = VIRT_IOMMU_NONE;
2621 } else {
2622 error_setg(errp, "Invalid iommu value");
2623 error_append_hint(errp, "Valid values are none, smmuv3.\n");
2624 }
2625 }
2626
2627 static bool virt_get_default_bus_bypass_iommu(Object *obj, Error **errp)
2628 {
2629 VirtMachineState *vms = VIRT_MACHINE(obj);
2630
2631 return vms->default_bus_bypass_iommu;
2632 }
2633
2634 static void virt_set_default_bus_bypass_iommu(Object *obj, bool value,
2635 Error **errp)
2636 {
2637 VirtMachineState *vms = VIRT_MACHINE(obj);
2638
2639 vms->default_bus_bypass_iommu = value;
2640 }
2641
2642 static CpuInstanceProperties
2643 virt_cpu_index_to_props(MachineState *ms, unsigned cpu_index)
2644 {
2645 MachineClass *mc = MACHINE_GET_CLASS(ms);
2646 const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms);
2647
2648 assert(cpu_index < possible_cpus->len);
2649 return possible_cpus->cpus[cpu_index].props;
2650 }
2651
2652 static int64_t virt_get_default_cpu_node_id(const MachineState *ms, int idx)
2653 {
2654 int64_t socket_id = ms->possible_cpus->cpus[idx].props.socket_id;
2655
2656 return socket_id % ms->numa_state->num_nodes;
2657 }
2658
2659 static const CPUArchIdList *virt_possible_cpu_arch_ids(MachineState *ms)
2660 {
2661 int n;
2662 unsigned int max_cpus = ms->smp.max_cpus;
2663 VirtMachineState *vms = VIRT_MACHINE(ms);
2664 MachineClass *mc = MACHINE_GET_CLASS(vms);
2665
2666 if (ms->possible_cpus) {
2667 assert(ms->possible_cpus->len == max_cpus);
2668 return ms->possible_cpus;
2669 }
2670
2671 ms->possible_cpus = g_malloc0(sizeof(CPUArchIdList) +
2672 sizeof(CPUArchId) * max_cpus);
2673 ms->possible_cpus->len = max_cpus;
2674 for (n = 0; n < ms->possible_cpus->len; n++) {
2675 ms->possible_cpus->cpus[n].type = ms->cpu_type;
2676 ms->possible_cpus->cpus[n].arch_id =
2677 virt_cpu_mp_affinity(vms, n);
2678
2679 assert(!mc->smp_props.dies_supported);
2680 ms->possible_cpus->cpus[n].props.has_socket_id = true;
2681 ms->possible_cpus->cpus[n].props.socket_id =
2682 n / (ms->smp.clusters * ms->smp.cores * ms->smp.threads);
2683 ms->possible_cpus->cpus[n].props.has_cluster_id = true;
2684 ms->possible_cpus->cpus[n].props.cluster_id =
2685 (n / (ms->smp.cores * ms->smp.threads)) % ms->smp.clusters;
2686 ms->possible_cpus->cpus[n].props.has_core_id = true;
2687 ms->possible_cpus->cpus[n].props.core_id =
2688 (n / ms->smp.threads) % ms->smp.cores;
2689 ms->possible_cpus->cpus[n].props.has_thread_id = true;
2690 ms->possible_cpus->cpus[n].props.thread_id =
2691 n % ms->smp.threads;
2692 }
2693 return ms->possible_cpus;
2694 }
2695
2696 static void virt_memory_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
2697 Error **errp)
2698 {
2699 VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2700 const MachineState *ms = MACHINE(hotplug_dev);
2701 const bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2702
2703 if (!vms->acpi_dev) {
2704 error_setg(errp,
2705 "memory hotplug is not enabled: missing acpi-ged device");
2706 return;
2707 }
2708
2709 if (vms->mte) {
2710 error_setg(errp, "memory hotplug is not enabled: MTE is enabled");
2711 return;
2712 }
2713
2714 if (is_nvdimm && !ms->nvdimms_state->is_enabled) {
2715 error_setg(errp, "nvdimm is not enabled: add 'nvdimm=on' to '-M'");
2716 return;
2717 }
2718
2719 pc_dimm_pre_plug(PC_DIMM(dev), MACHINE(hotplug_dev), NULL, errp);
2720 }
2721
2722 static void virt_memory_plug(HotplugHandler *hotplug_dev,
2723 DeviceState *dev, Error **errp)
2724 {
2725 VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2726 MachineState *ms = MACHINE(hotplug_dev);
2727 bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2728
2729 pc_dimm_plug(PC_DIMM(dev), MACHINE(vms));
2730
2731 if (is_nvdimm) {
2732 nvdimm_plug(ms->nvdimms_state);
2733 }
2734
2735 hotplug_handler_plug(HOTPLUG_HANDLER(vms->acpi_dev),
2736 dev, &error_abort);
2737 }
2738
2739 static void virt_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev,
2740 DeviceState *dev, Error **errp)
2741 {
2742 VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2743
2744 if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2745 virt_memory_pre_plug(hotplug_dev, dev, errp);
2746 } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI)) {
2747 virtio_md_pci_pre_plug(VIRTIO_MD_PCI(dev), MACHINE(hotplug_dev), errp);
2748 } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2749 hwaddr db_start = 0, db_end = 0;
2750 char *resv_prop_str;
2751
2752 if (vms->iommu != VIRT_IOMMU_NONE) {
2753 error_setg(errp, "virt machine does not support multiple IOMMUs");
2754 return;
2755 }
2756
2757 switch (vms->msi_controller) {
2758 case VIRT_MSI_CTRL_NONE:
2759 return;
2760 case VIRT_MSI_CTRL_ITS:
2761 /* GITS_TRANSLATER page */
2762 db_start = base_memmap[VIRT_GIC_ITS].base + 0x10000;
2763 db_end = base_memmap[VIRT_GIC_ITS].base +
2764 base_memmap[VIRT_GIC_ITS].size - 1;
2765 break;
2766 case VIRT_MSI_CTRL_GICV2M:
2767 /* MSI_SETSPI_NS page */
2768 db_start = base_memmap[VIRT_GIC_V2M].base;
2769 db_end = db_start + base_memmap[VIRT_GIC_V2M].size - 1;
2770 break;
2771 }
2772 resv_prop_str = g_strdup_printf("0x%"PRIx64":0x%"PRIx64":%u",
2773 db_start, db_end,
2774 VIRTIO_IOMMU_RESV_MEM_T_MSI);
2775
2776 object_property_set_uint(OBJECT(dev), "len-reserved-regions", 1, errp);
2777 object_property_set_str(OBJECT(dev), "reserved-regions[0]",
2778 resv_prop_str, errp);
2779 g_free(resv_prop_str);
2780 }
2781 }
2782
2783 static void virt_machine_device_plug_cb(HotplugHandler *hotplug_dev,
2784 DeviceState *dev, Error **errp)
2785 {
2786 VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2787
2788 if (vms->platform_bus_dev) {
2789 MachineClass *mc = MACHINE_GET_CLASS(vms);
2790
2791 if (device_is_dynamic_sysbus(mc, dev)) {
2792 platform_bus_link_device(PLATFORM_BUS_DEVICE(vms->platform_bus_dev),
2793 SYS_BUS_DEVICE(dev));
2794 }
2795 }
2796
2797 if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2798 virt_memory_plug(hotplug_dev, dev, errp);
2799 } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI)) {
2800 virtio_md_pci_plug(VIRTIO_MD_PCI(dev), MACHINE(hotplug_dev), errp);
2801 }
2802
2803 if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2804 PCIDevice *pdev = PCI_DEVICE(dev);
2805
2806 vms->iommu = VIRT_IOMMU_VIRTIO;
2807 vms->virtio_iommu_bdf = pci_get_bdf(pdev);
2808 create_virtio_iommu_dt_bindings(vms);
2809 }
2810 }
2811
2812 static void virt_dimm_unplug_request(HotplugHandler *hotplug_dev,
2813 DeviceState *dev, Error **errp)
2814 {
2815 VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2816
2817 if (!vms->acpi_dev) {
2818 error_setg(errp,
2819 "memory hotplug is not enabled: missing acpi-ged device");
2820 return;
2821 }
2822
2823 if (object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM)) {
2824 error_setg(errp, "nvdimm device hot unplug is not supported yet.");
2825 return;
2826 }
2827
2828 hotplug_handler_unplug_request(HOTPLUG_HANDLER(vms->acpi_dev), dev,
2829 errp);
2830 }
2831
2832 static void virt_dimm_unplug(HotplugHandler *hotplug_dev,
2833 DeviceState *dev, Error **errp)
2834 {
2835 VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2836 Error *local_err = NULL;
2837
2838 hotplug_handler_unplug(HOTPLUG_HANDLER(vms->acpi_dev), dev, &local_err);
2839 if (local_err) {
2840 goto out;
2841 }
2842
2843 pc_dimm_unplug(PC_DIMM(dev), MACHINE(vms));
2844 qdev_unrealize(dev);
2845
2846 out:
2847 error_propagate(errp, local_err);
2848 }
2849
2850 static void virt_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
2851 DeviceState *dev, Error **errp)
2852 {
2853 if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2854 virt_dimm_unplug_request(hotplug_dev, dev, errp);
2855 } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI)) {
2856 virtio_md_pci_unplug_request(VIRTIO_MD_PCI(dev), MACHINE(hotplug_dev),
2857 errp);
2858 } else {
2859 error_setg(errp, "device unplug request for unsupported device"
2860 " type: %s", object_get_typename(OBJECT(dev)));
2861 }
2862 }
2863
2864 static void virt_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
2865 DeviceState *dev, Error **errp)
2866 {
2867 if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2868 virt_dimm_unplug(hotplug_dev, dev, errp);
2869 } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI)) {
2870 virtio_md_pci_unplug(VIRTIO_MD_PCI(dev), MACHINE(hotplug_dev), errp);
2871 } else {
2872 error_setg(errp, "virt: device unplug for unsupported device"
2873 " type: %s", object_get_typename(OBJECT(dev)));
2874 }
2875 }
2876
2877 static HotplugHandler *virt_machine_get_hotplug_handler(MachineState *machine,
2878 DeviceState *dev)
2879 {
2880 MachineClass *mc = MACHINE_GET_CLASS(machine);
2881
2882 if (device_is_dynamic_sysbus(mc, dev) ||
2883 object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) ||
2884 object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI) ||
2885 object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2886 return HOTPLUG_HANDLER(machine);
2887 }
2888 return NULL;
2889 }
2890
2891 /*
2892 * for arm64 kvm_type [7-0] encodes the requested number of bits
2893 * in the IPA address space
2894 */
2895 static int virt_kvm_type(MachineState *ms, const char *type_str)
2896 {
2897 VirtMachineState *vms = VIRT_MACHINE(ms);
2898 int max_vm_pa_size, requested_pa_size;
2899 bool fixed_ipa;
2900
2901 max_vm_pa_size = kvm_arm_get_max_vm_ipa_size(ms, &fixed_ipa);
2902
2903 /* we freeze the memory map to compute the highest gpa */
2904 virt_set_memmap(vms, max_vm_pa_size);
2905
2906 requested_pa_size = 64 - clz64(vms->highest_gpa);
2907
2908 /*
2909 * KVM requires the IPA size to be at least 32 bits.
2910 */
2911 if (requested_pa_size < 32) {
2912 requested_pa_size = 32;
2913 }
2914
2915 if (requested_pa_size > max_vm_pa_size) {
2916 error_report("-m and ,maxmem option values "
2917 "require an IPA range (%d bits) larger than "
2918 "the one supported by the host (%d bits)",
2919 requested_pa_size, max_vm_pa_size);
2920 return -1;
2921 }
2922 /*
2923 * We return the requested PA log size, unless KVM only supports
2924 * the implicit legacy 40b IPA setting, in which case the kvm_type
2925 * must be 0.
2926 */
2927 return fixed_ipa ? 0 : requested_pa_size;
2928 }
2929
2930 static void virt_machine_class_init(ObjectClass *oc, void *data)
2931 {
2932 MachineClass *mc = MACHINE_CLASS(oc);
2933 HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
2934
2935 mc->init = machvirt_init;
2936 /* Start with max_cpus set to 512, which is the maximum supported by KVM.
2937 * The value may be reduced later when we have more information about the
2938 * configuration of the particular instance.
2939 */
2940 mc->max_cpus = 512;
2941 machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_CALXEDA_XGMAC);
2942 machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_AMD_XGBE);
2943 machine_class_allow_dynamic_sysbus_dev(mc, TYPE_RAMFB_DEVICE);
2944 machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_PLATFORM);
2945 #ifdef CONFIG_TPM
2946 machine_class_allow_dynamic_sysbus_dev(mc, TYPE_TPM_TIS_SYSBUS);
2947 #endif
2948 mc->block_default_type = IF_VIRTIO;
2949 mc->no_cdrom = 1;
2950 mc->pci_allow_0_address = true;
2951 /* We know we will never create a pre-ARMv7 CPU which needs 1K pages */
2952 mc->minimum_page_bits = 12;
2953 mc->possible_cpu_arch_ids = virt_possible_cpu_arch_ids;
2954 mc->cpu_index_to_instance_props = virt_cpu_index_to_props;
2955 #ifdef CONFIG_TCG
2956 mc->default_cpu_type = ARM_CPU_TYPE_NAME("cortex-a15");
2957 #else
2958 mc->default_cpu_type = ARM_CPU_TYPE_NAME("max");
2959 #endif
2960 mc->get_default_cpu_node_id = virt_get_default_cpu_node_id;
2961 mc->kvm_type = virt_kvm_type;
2962 assert(!mc->get_hotplug_handler);
2963 mc->get_hotplug_handler = virt_machine_get_hotplug_handler;
2964 hc->pre_plug = virt_machine_device_pre_plug_cb;
2965 hc->plug = virt_machine_device_plug_cb;
2966 hc->unplug_request = virt_machine_device_unplug_request_cb;
2967 hc->unplug = virt_machine_device_unplug_cb;
2968 mc->nvdimm_supported = true;
2969 mc->smp_props.clusters_supported = true;
2970 mc->auto_enable_numa_with_memhp = true;
2971 mc->auto_enable_numa_with_memdev = true;
2972 /* platform instead of architectural choice */
2973 mc->cpu_cluster_has_numa_boundary = true;
2974 mc->default_ram_id = "mach-virt.ram";
2975 mc->default_nic = "virtio-net-pci";
2976
2977 object_class_property_add(oc, "acpi", "OnOffAuto",
2978 virt_get_acpi, virt_set_acpi,
2979 NULL, NULL);
2980 object_class_property_set_description(oc, "acpi",
2981 "Enable ACPI");
2982 object_class_property_add_bool(oc, "secure", virt_get_secure,
2983 virt_set_secure);
2984 object_class_property_set_description(oc, "secure",
2985 "Set on/off to enable/disable the ARM "
2986 "Security Extensions (TrustZone)");
2987
2988 object_class_property_add_bool(oc, "virtualization", virt_get_virt,
2989 virt_set_virt);
2990 object_class_property_set_description(oc, "virtualization",
2991 "Set on/off to enable/disable emulating a "
2992 "guest CPU which implements the ARM "
2993 "Virtualization Extensions");
2994
2995 object_class_property_add_bool(oc, "highmem", virt_get_highmem,
2996 virt_set_highmem);
2997 object_class_property_set_description(oc, "highmem",
2998 "Set on/off to enable/disable using "
2999 "physical address space above 32 bits");
3000
3001 object_class_property_add_bool(oc, "compact-highmem",
3002 virt_get_compact_highmem,
3003 virt_set_compact_highmem);
3004 object_class_property_set_description(oc, "compact-highmem",
3005 "Set on/off to enable/disable compact "
3006 "layout for high memory regions");
3007
3008 object_class_property_add_bool(oc, "highmem-redists",
3009 virt_get_highmem_redists,
3010 virt_set_highmem_redists);
3011 object_class_property_set_description(oc, "highmem-redists",
3012 "Set on/off to enable/disable high "
3013 "memory region for GICv3 or GICv4 "
3014 "redistributor");
3015
3016 object_class_property_add_bool(oc, "highmem-ecam",
3017 virt_get_highmem_ecam,
3018 virt_set_highmem_ecam);
3019 object_class_property_set_description(oc, "highmem-ecam",
3020 "Set on/off to enable/disable high "
3021 "memory region for PCI ECAM");
3022
3023 object_class_property_add_bool(oc, "highmem-mmio",
3024 virt_get_highmem_mmio,
3025 virt_set_highmem_mmio);
3026 object_class_property_set_description(oc, "highmem-mmio",
3027 "Set on/off to enable/disable high "
3028 "memory region for PCI MMIO");
3029
3030 object_class_property_add_str(oc, "gic-version", virt_get_gic_version,
3031 virt_set_gic_version);
3032 object_class_property_set_description(oc, "gic-version",
3033 "Set GIC version. "
3034 "Valid values are 2, 3, 4, host and max");
3035
3036 object_class_property_add_str(oc, "iommu", virt_get_iommu, virt_set_iommu);
3037 object_class_property_set_description(oc, "iommu",
3038 "Set the IOMMU type. "
3039 "Valid values are none and smmuv3");
3040
3041 object_class_property_add_bool(oc, "default-bus-bypass-iommu",
3042 virt_get_default_bus_bypass_iommu,
3043 virt_set_default_bus_bypass_iommu);
3044 object_class_property_set_description(oc, "default-bus-bypass-iommu",
3045 "Set on/off to enable/disable "
3046 "bypass_iommu for default root bus");
3047
3048 object_class_property_add_bool(oc, "ras", virt_get_ras,
3049 virt_set_ras);
3050 object_class_property_set_description(oc, "ras",
3051 "Set on/off to enable/disable reporting host memory errors "
3052 "to a KVM guest using ACPI and guest external abort exceptions");
3053
3054 object_class_property_add_bool(oc, "mte", virt_get_mte, virt_set_mte);
3055 object_class_property_set_description(oc, "mte",
3056 "Set on/off to enable/disable emulating a "
3057 "guest CPU which implements the ARM "
3058 "Memory Tagging Extension");
3059
3060 object_class_property_add_bool(oc, "its", virt_get_its,
3061 virt_set_its);
3062 object_class_property_set_description(oc, "its",
3063 "Set on/off to enable/disable "
3064 "ITS instantiation");
3065
3066 object_class_property_add_bool(oc, "dtb-randomness",
3067 virt_get_dtb_randomness,
3068 virt_set_dtb_randomness);
3069 object_class_property_set_description(oc, "dtb-randomness",
3070 "Set off to disable passing random or "
3071 "non-deterministic dtb nodes to guest");
3072
3073 object_class_property_add_bool(oc, "dtb-kaslr-seed",
3074 virt_get_dtb_randomness,
3075 virt_set_dtb_randomness);
3076 object_class_property_set_description(oc, "dtb-kaslr-seed",
3077 "Deprecated synonym of dtb-randomness");
3078
3079 object_class_property_add_str(oc, "x-oem-id",
3080 virt_get_oem_id,
3081 virt_set_oem_id);
3082 object_class_property_set_description(oc, "x-oem-id",
3083 "Override the default value of field OEMID "
3084 "in ACPI table header."
3085 "The string may be up to 6 bytes in size");
3086
3087
3088 object_class_property_add_str(oc, "x-oem-table-id",
3089 virt_get_oem_table_id,
3090 virt_set_oem_table_id);
3091 object_class_property_set_description(oc, "x-oem-table-id",
3092 "Override the default value of field OEM Table ID "
3093 "in ACPI table header."
3094 "The string may be up to 8 bytes in size");
3095
3096 }
3097
3098 static void virt_instance_init(Object *obj)
3099 {
3100 VirtMachineState *vms = VIRT_MACHINE(obj);
3101 VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
3102
3103 /* EL3 is disabled by default on virt: this makes us consistent
3104 * between KVM and TCG for this board, and it also allows us to
3105 * boot UEFI blobs which assume no TrustZone support.
3106 */
3107 vms->secure = false;
3108
3109 /* EL2 is also disabled by default, for similar reasons */
3110 vms->virt = false;
3111
3112 /* High memory is enabled by default */
3113 vms->highmem = true;
3114 vms->highmem_compact = !vmc->no_highmem_compact;
3115 vms->gic_version = VIRT_GIC_VERSION_NOSEL;
3116
3117 vms->highmem_ecam = !vmc->no_highmem_ecam;
3118 vms->highmem_mmio = true;
3119 vms->highmem_redists = true;
3120
3121 if (vmc->no_its) {
3122 vms->its = false;
3123 } else {
3124 /* Default allows ITS instantiation */
3125 vms->its = true;
3126
3127 if (vmc->no_tcg_its) {
3128 vms->tcg_its = false;
3129 } else {
3130 vms->tcg_its = true;
3131 }
3132 }
3133
3134 /* Default disallows iommu instantiation */
3135 vms->iommu = VIRT_IOMMU_NONE;
3136
3137 /* The default root bus is attached to iommu by default */
3138 vms->default_bus_bypass_iommu = false;
3139
3140 /* Default disallows RAS instantiation */
3141 vms->ras = false;
3142
3143 /* MTE is disabled by default. */
3144 vms->mte = false;
3145
3146 /* Supply kaslr-seed and rng-seed by default */
3147 vms->dtb_randomness = true;
3148
3149 vms->irqmap = a15irqmap;
3150
3151 virt_flash_create(vms);
3152
3153 vms->oem_id = g_strndup(ACPI_BUILD_APPNAME6, 6);
3154 vms->oem_table_id = g_strndup(ACPI_BUILD_APPNAME8, 8);
3155 }
3156
3157 static const TypeInfo virt_machine_info = {
3158 .name = TYPE_VIRT_MACHINE,
3159 .parent = TYPE_MACHINE,
3160 .abstract = true,
3161 .instance_size = sizeof(VirtMachineState),
3162 .class_size = sizeof(VirtMachineClass),
3163 .class_init = virt_machine_class_init,
3164 .instance_init = virt_instance_init,
3165 .interfaces = (InterfaceInfo[]) {
3166 { TYPE_HOTPLUG_HANDLER },
3167 { }
3168 },
3169 };
3170
3171 static void machvirt_machine_init(void)
3172 {
3173 type_register_static(&virt_machine_info);
3174 }
3175 type_init(machvirt_machine_init);
3176
3177 static void virt_machine_8_2_options(MachineClass *mc)
3178 {
3179 }
3180 DEFINE_VIRT_MACHINE_AS_LATEST(8, 2)
3181
3182 static void virt_machine_8_1_options(MachineClass *mc)
3183 {
3184 virt_machine_8_2_options(mc);
3185 compat_props_add(mc->compat_props, hw_compat_8_1, hw_compat_8_1_len);
3186 }
3187 DEFINE_VIRT_MACHINE(8, 1)
3188
3189 static void virt_machine_8_0_options(MachineClass *mc)
3190 {
3191 virt_machine_8_1_options(mc);
3192 compat_props_add(mc->compat_props, hw_compat_8_0, hw_compat_8_0_len);
3193 }
3194 DEFINE_VIRT_MACHINE(8, 0)
3195
3196 static void virt_machine_7_2_options(MachineClass *mc)
3197 {
3198 virt_machine_8_0_options(mc);
3199 compat_props_add(mc->compat_props, hw_compat_7_2, hw_compat_7_2_len);
3200 }
3201 DEFINE_VIRT_MACHINE(7, 2)
3202
3203 static void virt_machine_7_1_options(MachineClass *mc)
3204 {
3205 VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3206
3207 virt_machine_7_2_options(mc);
3208 compat_props_add(mc->compat_props, hw_compat_7_1, hw_compat_7_1_len);
3209 /* Compact layout for high memory regions was introduced with 7.2 */
3210 vmc->no_highmem_compact = true;
3211 }
3212 DEFINE_VIRT_MACHINE(7, 1)
3213
3214 static void virt_machine_7_0_options(MachineClass *mc)
3215 {
3216 virt_machine_7_1_options(mc);
3217 compat_props_add(mc->compat_props, hw_compat_7_0, hw_compat_7_0_len);
3218 }
3219 DEFINE_VIRT_MACHINE(7, 0)
3220
3221 static void virt_machine_6_2_options(MachineClass *mc)
3222 {
3223 VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3224
3225 virt_machine_7_0_options(mc);
3226 compat_props_add(mc->compat_props, hw_compat_6_2, hw_compat_6_2_len);
3227 vmc->no_tcg_lpa2 = true;
3228 }
3229 DEFINE_VIRT_MACHINE(6, 2)
3230
3231 static void virt_machine_6_1_options(MachineClass *mc)
3232 {
3233 VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3234
3235 virt_machine_6_2_options(mc);
3236 compat_props_add(mc->compat_props, hw_compat_6_1, hw_compat_6_1_len);
3237 mc->smp_props.prefer_sockets = true;
3238 vmc->no_cpu_topology = true;
3239
3240 /* qemu ITS was introduced with 6.2 */
3241 vmc->no_tcg_its = true;
3242 }
3243 DEFINE_VIRT_MACHINE(6, 1)
3244
3245 static void virt_machine_6_0_options(MachineClass *mc)
3246 {
3247 virt_machine_6_1_options(mc);
3248 compat_props_add(mc->compat_props, hw_compat_6_0, hw_compat_6_0_len);
3249 }
3250 DEFINE_VIRT_MACHINE(6, 0)
3251
3252 static void virt_machine_5_2_options(MachineClass *mc)
3253 {
3254 VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3255
3256 virt_machine_6_0_options(mc);
3257 compat_props_add(mc->compat_props, hw_compat_5_2, hw_compat_5_2_len);
3258 vmc->no_secure_gpio = true;
3259 }
3260 DEFINE_VIRT_MACHINE(5, 2)
3261
3262 static void virt_machine_5_1_options(MachineClass *mc)
3263 {
3264 VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3265
3266 virt_machine_5_2_options(mc);
3267 compat_props_add(mc->compat_props, hw_compat_5_1, hw_compat_5_1_len);
3268 vmc->no_kvm_steal_time = true;
3269 }
3270 DEFINE_VIRT_MACHINE(5, 1)
3271
3272 static void virt_machine_5_0_options(MachineClass *mc)
3273 {
3274 VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3275
3276 virt_machine_5_1_options(mc);
3277 compat_props_add(mc->compat_props, hw_compat_5_0, hw_compat_5_0_len);
3278 mc->numa_mem_supported = true;
3279 vmc->acpi_expose_flash = true;
3280 mc->auto_enable_numa_with_memdev = false;
3281 }
3282 DEFINE_VIRT_MACHINE(5, 0)
3283
3284 static void virt_machine_4_2_options(MachineClass *mc)
3285 {
3286 VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3287
3288 virt_machine_5_0_options(mc);
3289 compat_props_add(mc->compat_props, hw_compat_4_2, hw_compat_4_2_len);
3290 vmc->kvm_no_adjvtime = true;
3291 }
3292 DEFINE_VIRT_MACHINE(4, 2)
3293
3294 static void virt_machine_4_1_options(MachineClass *mc)
3295 {
3296 VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3297
3298 virt_machine_4_2_options(mc);
3299 compat_props_add(mc->compat_props, hw_compat_4_1, hw_compat_4_1_len);
3300 vmc->no_ged = true;
3301 mc->auto_enable_numa_with_memhp = false;
3302 }
3303 DEFINE_VIRT_MACHINE(4, 1)
3304
3305 static void virt_machine_4_0_options(MachineClass *mc)
3306 {
3307 virt_machine_4_1_options(mc);
3308 compat_props_add(mc->compat_props, hw_compat_4_0, hw_compat_4_0_len);
3309 }
3310 DEFINE_VIRT_MACHINE(4, 0)
3311
3312 static void virt_machine_3_1_options(MachineClass *mc)
3313 {
3314 virt_machine_4_0_options(mc);
3315 compat_props_add(mc->compat_props, hw_compat_3_1, hw_compat_3_1_len);
3316 }
3317 DEFINE_VIRT_MACHINE(3, 1)
3318
3319 static void virt_machine_3_0_options(MachineClass *mc)
3320 {
3321 virt_machine_3_1_options(mc);
3322 compat_props_add(mc->compat_props, hw_compat_3_0, hw_compat_3_0_len);
3323 }
3324 DEFINE_VIRT_MACHINE(3, 0)
3325
3326 static void virt_machine_2_12_options(MachineClass *mc)
3327 {
3328 VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3329
3330 virt_machine_3_0_options(mc);
3331 compat_props_add(mc->compat_props, hw_compat_2_12, hw_compat_2_12_len);
3332 vmc->no_highmem_ecam = true;
3333 mc->max_cpus = 255;
3334 }
3335 DEFINE_VIRT_MACHINE(2, 12)
3336
3337 static void virt_machine_2_11_options(MachineClass *mc)
3338 {
3339 VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3340
3341 virt_machine_2_12_options(mc);
3342 compat_props_add(mc->compat_props, hw_compat_2_11, hw_compat_2_11_len);
3343 vmc->smbios_old_sys_ver = true;
3344 }
3345 DEFINE_VIRT_MACHINE(2, 11)
3346
3347 static void virt_machine_2_10_options(MachineClass *mc)
3348 {
3349 virt_machine_2_11_options(mc);
3350 compat_props_add(mc->compat_props, hw_compat_2_10, hw_compat_2_10_len);
3351 /* before 2.11 we never faulted accesses to bad addresses */
3352 mc->ignore_memory_transaction_failures = true;
3353 }
3354 DEFINE_VIRT_MACHINE(2, 10)
3355
3356 static void virt_machine_2_9_options(MachineClass *mc)
3357 {
3358 virt_machine_2_10_options(mc);
3359 compat_props_add(mc->compat_props, hw_compat_2_9, hw_compat_2_9_len);
3360 }
3361 DEFINE_VIRT_MACHINE(2, 9)
3362
3363 static void virt_machine_2_8_options(MachineClass *mc)
3364 {
3365 VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3366
3367 virt_machine_2_9_options(mc);
3368 compat_props_add(mc->compat_props, hw_compat_2_8, hw_compat_2_8_len);
3369 /* For 2.8 and earlier we falsely claimed in the DT that
3370 * our timers were edge-triggered, not level-triggered.
3371 */
3372 vmc->claim_edge_triggered_timers = true;
3373 }
3374 DEFINE_VIRT_MACHINE(2, 8)
3375
3376 static void virt_machine_2_7_options(MachineClass *mc)
3377 {
3378 VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3379
3380 virt_machine_2_8_options(mc);
3381 compat_props_add(mc->compat_props, hw_compat_2_7, hw_compat_2_7_len);
3382 /* ITS was introduced with 2.8 */
3383 vmc->no_its = true;
3384 /* Stick with 1K pages for migration compatibility */
3385 mc->minimum_page_bits = 0;
3386 }
3387 DEFINE_VIRT_MACHINE(2, 7)
3388
3389 static void virt_machine_2_6_options(MachineClass *mc)
3390 {
3391 VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3392
3393 virt_machine_2_7_options(mc);
3394 compat_props_add(mc->compat_props, hw_compat_2_6, hw_compat_2_6_len);
3395 vmc->disallow_affinity_adjustment = true;
3396 /* Disable PMU for 2.6 as PMU support was first introduced in 2.7 */
3397 vmc->no_pmu = true;
3398 }
3399 DEFINE_VIRT_MACHINE(2, 6)