]> git.proxmox.com Git - mirror_qemu.git/blob - hw/arm/virt.c
Merge remote-tracking branch 'remotes/kraxel/tags/input-20180618-pull-request' into...
[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 "qapi/error.h"
33 #include "hw/sysbus.h"
34 #include "hw/arm/arm.h"
35 #include "hw/arm/primecell.h"
36 #include "hw/arm/virt.h"
37 #include "hw/vfio/vfio-calxeda-xgmac.h"
38 #include "hw/vfio/vfio-amd-xgbe.h"
39 #include "hw/display/ramfb.h"
40 #include "hw/devices.h"
41 #include "net/net.h"
42 #include "sysemu/device_tree.h"
43 #include "sysemu/numa.h"
44 #include "sysemu/sysemu.h"
45 #include "sysemu/kvm.h"
46 #include "hw/compat.h"
47 #include "hw/loader.h"
48 #include "exec/address-spaces.h"
49 #include "qemu/bitops.h"
50 #include "qemu/error-report.h"
51 #include "hw/pci-host/gpex.h"
52 #include "hw/arm/sysbus-fdt.h"
53 #include "hw/platform-bus.h"
54 #include "hw/arm/fdt.h"
55 #include "hw/intc/arm_gic.h"
56 #include "hw/intc/arm_gicv3_common.h"
57 #include "kvm_arm.h"
58 #include "hw/smbios/smbios.h"
59 #include "qapi/visitor.h"
60 #include "standard-headers/linux/input.h"
61 #include "hw/arm/smmuv3.h"
62
63 #define DEFINE_VIRT_MACHINE_LATEST(major, minor, latest) \
64 static void virt_##major##_##minor##_class_init(ObjectClass *oc, \
65 void *data) \
66 { \
67 MachineClass *mc = MACHINE_CLASS(oc); \
68 virt_machine_##major##_##minor##_options(mc); \
69 mc->desc = "QEMU " # major "." # minor " ARM Virtual Machine"; \
70 if (latest) { \
71 mc->alias = "virt"; \
72 } \
73 } \
74 static const TypeInfo machvirt_##major##_##minor##_info = { \
75 .name = MACHINE_TYPE_NAME("virt-" # major "." # minor), \
76 .parent = TYPE_VIRT_MACHINE, \
77 .instance_init = virt_##major##_##minor##_instance_init, \
78 .class_init = virt_##major##_##minor##_class_init, \
79 }; \
80 static void machvirt_machine_##major##_##minor##_init(void) \
81 { \
82 type_register_static(&machvirt_##major##_##minor##_info); \
83 } \
84 type_init(machvirt_machine_##major##_##minor##_init);
85
86 #define DEFINE_VIRT_MACHINE_AS_LATEST(major, minor) \
87 DEFINE_VIRT_MACHINE_LATEST(major, minor, true)
88 #define DEFINE_VIRT_MACHINE(major, minor) \
89 DEFINE_VIRT_MACHINE_LATEST(major, minor, false)
90
91
92 /* Number of external interrupt lines to configure the GIC with */
93 #define NUM_IRQS 256
94
95 #define PLATFORM_BUS_NUM_IRQS 64
96
97 /* RAM limit in GB. Since VIRT_MEM starts at the 1GB mark, this means
98 * RAM can go up to the 256GB mark, leaving 256GB of the physical
99 * address space unallocated and free for future use between 256G and 512G.
100 * If we need to provide more RAM to VMs in the future then we need to:
101 * * allocate a second bank of RAM starting at 2TB and working up
102 * * fix the DT and ACPI table generation code in QEMU to correctly
103 * report two split lumps of RAM to the guest
104 * * fix KVM in the host kernel to allow guests with >40 bit address spaces
105 * (We don't want to fill all the way up to 512GB with RAM because
106 * we might want it for non-RAM purposes later. Conversely it seems
107 * reasonable to assume that anybody configuring a VM with a quarter
108 * of a terabyte of RAM will be doing it on a host with more than a
109 * terabyte of physical address space.)
110 */
111 #define RAMLIMIT_GB 255
112 #define RAMLIMIT_BYTES (RAMLIMIT_GB * 1024ULL * 1024 * 1024)
113
114 /* Addresses and sizes of our components.
115 * 0..128MB is space for a flash device so we can run bootrom code such as UEFI.
116 * 128MB..256MB is used for miscellaneous device I/O.
117 * 256MB..1GB is reserved for possible future PCI support (ie where the
118 * PCI memory window will go if we add a PCI host controller).
119 * 1GB and up is RAM (which may happily spill over into the
120 * high memory region beyond 4GB).
121 * This represents a compromise between how much RAM can be given to
122 * a 32 bit VM and leaving space for expansion and in particular for PCI.
123 * Note that devices should generally be placed at multiples of 0x10000,
124 * to accommodate guests using 64K pages.
125 */
126 static const MemMapEntry a15memmap[] = {
127 /* Space up to 0x8000000 is reserved for a boot ROM */
128 [VIRT_FLASH] = { 0, 0x08000000 },
129 [VIRT_CPUPERIPHS] = { 0x08000000, 0x00020000 },
130 /* GIC distributor and CPU interfaces sit inside the CPU peripheral space */
131 [VIRT_GIC_DIST] = { 0x08000000, 0x00010000 },
132 [VIRT_GIC_CPU] = { 0x08010000, 0x00010000 },
133 [VIRT_GIC_V2M] = { 0x08020000, 0x00001000 },
134 /* The space in between here is reserved for GICv3 CPU/vCPU/HYP */
135 [VIRT_GIC_ITS] = { 0x08080000, 0x00020000 },
136 /* This redistributor space allows up to 2*64kB*123 CPUs */
137 [VIRT_GIC_REDIST] = { 0x080A0000, 0x00F60000 },
138 [VIRT_UART] = { 0x09000000, 0x00001000 },
139 [VIRT_RTC] = { 0x09010000, 0x00001000 },
140 [VIRT_FW_CFG] = { 0x09020000, 0x00000018 },
141 [VIRT_GPIO] = { 0x09030000, 0x00001000 },
142 [VIRT_SECURE_UART] = { 0x09040000, 0x00001000 },
143 [VIRT_SMMU] = { 0x09050000, 0x00020000 },
144 [VIRT_MMIO] = { 0x0a000000, 0x00000200 },
145 /* ...repeating for a total of NUM_VIRTIO_TRANSPORTS, each of that size */
146 [VIRT_PLATFORM_BUS] = { 0x0c000000, 0x02000000 },
147 [VIRT_SECURE_MEM] = { 0x0e000000, 0x01000000 },
148 [VIRT_PCIE_MMIO] = { 0x10000000, 0x2eff0000 },
149 [VIRT_PCIE_PIO] = { 0x3eff0000, 0x00010000 },
150 [VIRT_PCIE_ECAM] = { 0x3f000000, 0x01000000 },
151 [VIRT_MEM] = { 0x40000000, RAMLIMIT_BYTES },
152 /* Second PCIe window, 512GB wide at the 512GB boundary */
153 [VIRT_PCIE_MMIO_HIGH] = { 0x8000000000ULL, 0x8000000000ULL },
154 };
155
156 static const int a15irqmap[] = {
157 [VIRT_UART] = 1,
158 [VIRT_RTC] = 2,
159 [VIRT_PCIE] = 3, /* ... to 6 */
160 [VIRT_GPIO] = 7,
161 [VIRT_SECURE_UART] = 8,
162 [VIRT_MMIO] = 16, /* ...to 16 + NUM_VIRTIO_TRANSPORTS - 1 */
163 [VIRT_GIC_V2M] = 48, /* ...to 48 + NUM_GICV2M_SPIS - 1 */
164 [VIRT_SMMU] = 74, /* ...to 74 + NUM_SMMU_IRQS - 1 */
165 [VIRT_PLATFORM_BUS] = 112, /* ...to 112 + PLATFORM_BUS_NUM_IRQS -1 */
166 };
167
168 static const char *valid_cpus[] = {
169 ARM_CPU_TYPE_NAME("cortex-a15"),
170 ARM_CPU_TYPE_NAME("cortex-a53"),
171 ARM_CPU_TYPE_NAME("cortex-a57"),
172 ARM_CPU_TYPE_NAME("host"),
173 ARM_CPU_TYPE_NAME("max"),
174 };
175
176 static bool cpu_type_valid(const char *cpu)
177 {
178 int i;
179
180 for (i = 0; i < ARRAY_SIZE(valid_cpus); i++) {
181 if (strcmp(cpu, valid_cpus[i]) == 0) {
182 return true;
183 }
184 }
185 return false;
186 }
187
188 static void create_fdt(VirtMachineState *vms)
189 {
190 void *fdt = create_device_tree(&vms->fdt_size);
191
192 if (!fdt) {
193 error_report("create_device_tree() failed");
194 exit(1);
195 }
196
197 vms->fdt = fdt;
198
199 /* Header */
200 qemu_fdt_setprop_string(fdt, "/", "compatible", "linux,dummy-virt");
201 qemu_fdt_setprop_cell(fdt, "/", "#address-cells", 0x2);
202 qemu_fdt_setprop_cell(fdt, "/", "#size-cells", 0x2);
203
204 /*
205 * /chosen and /memory nodes must exist for load_dtb
206 * to fill in necessary properties later
207 */
208 qemu_fdt_add_subnode(fdt, "/chosen");
209 qemu_fdt_add_subnode(fdt, "/memory");
210 qemu_fdt_setprop_string(fdt, "/memory", "device_type", "memory");
211
212 /* Clock node, for the benefit of the UART. The kernel device tree
213 * binding documentation claims the PL011 node clock properties are
214 * optional but in practice if you omit them the kernel refuses to
215 * probe for the device.
216 */
217 vms->clock_phandle = qemu_fdt_alloc_phandle(fdt);
218 qemu_fdt_add_subnode(fdt, "/apb-pclk");
219 qemu_fdt_setprop_string(fdt, "/apb-pclk", "compatible", "fixed-clock");
220 qemu_fdt_setprop_cell(fdt, "/apb-pclk", "#clock-cells", 0x0);
221 qemu_fdt_setprop_cell(fdt, "/apb-pclk", "clock-frequency", 24000000);
222 qemu_fdt_setprop_string(fdt, "/apb-pclk", "clock-output-names",
223 "clk24mhz");
224 qemu_fdt_setprop_cell(fdt, "/apb-pclk", "phandle", vms->clock_phandle);
225
226 if (have_numa_distance) {
227 int size = nb_numa_nodes * nb_numa_nodes * 3 * sizeof(uint32_t);
228 uint32_t *matrix = g_malloc0(size);
229 int idx, i, j;
230
231 for (i = 0; i < nb_numa_nodes; i++) {
232 for (j = 0; j < nb_numa_nodes; j++) {
233 idx = (i * nb_numa_nodes + j) * 3;
234 matrix[idx + 0] = cpu_to_be32(i);
235 matrix[idx + 1] = cpu_to_be32(j);
236 matrix[idx + 2] = cpu_to_be32(numa_info[i].distance[j]);
237 }
238 }
239
240 qemu_fdt_add_subnode(fdt, "/distance-map");
241 qemu_fdt_setprop_string(fdt, "/distance-map", "compatible",
242 "numa-distance-map-v1");
243 qemu_fdt_setprop(fdt, "/distance-map", "distance-matrix",
244 matrix, size);
245 g_free(matrix);
246 }
247 }
248
249 static void fdt_add_timer_nodes(const VirtMachineState *vms)
250 {
251 /* On real hardware these interrupts are level-triggered.
252 * On KVM they were edge-triggered before host kernel version 4.4,
253 * and level-triggered afterwards.
254 * On emulated QEMU they are level-triggered.
255 *
256 * Getting the DTB info about them wrong is awkward for some
257 * guest kernels:
258 * pre-4.8 ignore the DT and leave the interrupt configured
259 * with whatever the GIC reset value (or the bootloader) left it at
260 * 4.8 before rc6 honour the incorrect data by programming it back
261 * into the GIC, causing problems
262 * 4.8rc6 and later ignore the DT and always write "level triggered"
263 * into the GIC
264 *
265 * For backwards-compatibility, virt-2.8 and earlier will continue
266 * to say these are edge-triggered, but later machines will report
267 * the correct information.
268 */
269 ARMCPU *armcpu;
270 VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
271 uint32_t irqflags = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
272
273 if (vmc->claim_edge_triggered_timers) {
274 irqflags = GIC_FDT_IRQ_FLAGS_EDGE_LO_HI;
275 }
276
277 if (vms->gic_version == 2) {
278 irqflags = deposit32(irqflags, GIC_FDT_IRQ_PPI_CPU_START,
279 GIC_FDT_IRQ_PPI_CPU_WIDTH,
280 (1 << vms->smp_cpus) - 1);
281 }
282
283 qemu_fdt_add_subnode(vms->fdt, "/timer");
284
285 armcpu = ARM_CPU(qemu_get_cpu(0));
286 if (arm_feature(&armcpu->env, ARM_FEATURE_V8)) {
287 const char compat[] = "arm,armv8-timer\0arm,armv7-timer";
288 qemu_fdt_setprop(vms->fdt, "/timer", "compatible",
289 compat, sizeof(compat));
290 } else {
291 qemu_fdt_setprop_string(vms->fdt, "/timer", "compatible",
292 "arm,armv7-timer");
293 }
294 qemu_fdt_setprop(vms->fdt, "/timer", "always-on", NULL, 0);
295 qemu_fdt_setprop_cells(vms->fdt, "/timer", "interrupts",
296 GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_S_EL1_IRQ, irqflags,
297 GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_NS_EL1_IRQ, irqflags,
298 GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_VIRT_IRQ, irqflags,
299 GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_NS_EL2_IRQ, irqflags);
300 }
301
302 static void fdt_add_cpu_nodes(const VirtMachineState *vms)
303 {
304 int cpu;
305 int addr_cells = 1;
306 const MachineState *ms = MACHINE(vms);
307
308 /*
309 * From Documentation/devicetree/bindings/arm/cpus.txt
310 * On ARM v8 64-bit systems value should be set to 2,
311 * that corresponds to the MPIDR_EL1 register size.
312 * If MPIDR_EL1[63:32] value is equal to 0 on all CPUs
313 * in the system, #address-cells can be set to 1, since
314 * MPIDR_EL1[63:32] bits are not used for CPUs
315 * identification.
316 *
317 * Here we actually don't know whether our system is 32- or 64-bit one.
318 * The simplest way to go is to examine affinity IDs of all our CPUs. If
319 * at least one of them has Aff3 populated, we set #address-cells to 2.
320 */
321 for (cpu = 0; cpu < vms->smp_cpus; cpu++) {
322 ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
323
324 if (armcpu->mp_affinity & ARM_AFF3_MASK) {
325 addr_cells = 2;
326 break;
327 }
328 }
329
330 qemu_fdt_add_subnode(vms->fdt, "/cpus");
331 qemu_fdt_setprop_cell(vms->fdt, "/cpus", "#address-cells", addr_cells);
332 qemu_fdt_setprop_cell(vms->fdt, "/cpus", "#size-cells", 0x0);
333
334 for (cpu = vms->smp_cpus - 1; cpu >= 0; cpu--) {
335 char *nodename = g_strdup_printf("/cpus/cpu@%d", cpu);
336 ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
337 CPUState *cs = CPU(armcpu);
338
339 qemu_fdt_add_subnode(vms->fdt, nodename);
340 qemu_fdt_setprop_string(vms->fdt, nodename, "device_type", "cpu");
341 qemu_fdt_setprop_string(vms->fdt, nodename, "compatible",
342 armcpu->dtb_compatible);
343
344 if (vms->psci_conduit != QEMU_PSCI_CONDUIT_DISABLED
345 && vms->smp_cpus > 1) {
346 qemu_fdt_setprop_string(vms->fdt, nodename,
347 "enable-method", "psci");
348 }
349
350 if (addr_cells == 2) {
351 qemu_fdt_setprop_u64(vms->fdt, nodename, "reg",
352 armcpu->mp_affinity);
353 } else {
354 qemu_fdt_setprop_cell(vms->fdt, nodename, "reg",
355 armcpu->mp_affinity);
356 }
357
358 if (ms->possible_cpus->cpus[cs->cpu_index].props.has_node_id) {
359 qemu_fdt_setprop_cell(vms->fdt, nodename, "numa-node-id",
360 ms->possible_cpus->cpus[cs->cpu_index].props.node_id);
361 }
362
363 g_free(nodename);
364 }
365 }
366
367 static void fdt_add_its_gic_node(VirtMachineState *vms)
368 {
369 vms->msi_phandle = qemu_fdt_alloc_phandle(vms->fdt);
370 qemu_fdt_add_subnode(vms->fdt, "/intc/its");
371 qemu_fdt_setprop_string(vms->fdt, "/intc/its", "compatible",
372 "arm,gic-v3-its");
373 qemu_fdt_setprop(vms->fdt, "/intc/its", "msi-controller", NULL, 0);
374 qemu_fdt_setprop_sized_cells(vms->fdt, "/intc/its", "reg",
375 2, vms->memmap[VIRT_GIC_ITS].base,
376 2, vms->memmap[VIRT_GIC_ITS].size);
377 qemu_fdt_setprop_cell(vms->fdt, "/intc/its", "phandle", vms->msi_phandle);
378 }
379
380 static void fdt_add_v2m_gic_node(VirtMachineState *vms)
381 {
382 vms->msi_phandle = qemu_fdt_alloc_phandle(vms->fdt);
383 qemu_fdt_add_subnode(vms->fdt, "/intc/v2m");
384 qemu_fdt_setprop_string(vms->fdt, "/intc/v2m", "compatible",
385 "arm,gic-v2m-frame");
386 qemu_fdt_setprop(vms->fdt, "/intc/v2m", "msi-controller", NULL, 0);
387 qemu_fdt_setprop_sized_cells(vms->fdt, "/intc/v2m", "reg",
388 2, vms->memmap[VIRT_GIC_V2M].base,
389 2, vms->memmap[VIRT_GIC_V2M].size);
390 qemu_fdt_setprop_cell(vms->fdt, "/intc/v2m", "phandle", vms->msi_phandle);
391 }
392
393 static void fdt_add_gic_node(VirtMachineState *vms)
394 {
395 vms->gic_phandle = qemu_fdt_alloc_phandle(vms->fdt);
396 qemu_fdt_setprop_cell(vms->fdt, "/", "interrupt-parent", vms->gic_phandle);
397
398 qemu_fdt_add_subnode(vms->fdt, "/intc");
399 qemu_fdt_setprop_cell(vms->fdt, "/intc", "#interrupt-cells", 3);
400 qemu_fdt_setprop(vms->fdt, "/intc", "interrupt-controller", NULL, 0);
401 qemu_fdt_setprop_cell(vms->fdt, "/intc", "#address-cells", 0x2);
402 qemu_fdt_setprop_cell(vms->fdt, "/intc", "#size-cells", 0x2);
403 qemu_fdt_setprop(vms->fdt, "/intc", "ranges", NULL, 0);
404 if (vms->gic_version == 3) {
405 qemu_fdt_setprop_string(vms->fdt, "/intc", "compatible",
406 "arm,gic-v3");
407 qemu_fdt_setprop_sized_cells(vms->fdt, "/intc", "reg",
408 2, vms->memmap[VIRT_GIC_DIST].base,
409 2, vms->memmap[VIRT_GIC_DIST].size,
410 2, vms->memmap[VIRT_GIC_REDIST].base,
411 2, vms->memmap[VIRT_GIC_REDIST].size);
412 if (vms->virt) {
413 qemu_fdt_setprop_cells(vms->fdt, "/intc", "interrupts",
414 GIC_FDT_IRQ_TYPE_PPI, ARCH_GICV3_MAINT_IRQ,
415 GIC_FDT_IRQ_FLAGS_LEVEL_HI);
416 }
417 } else {
418 /* 'cortex-a15-gic' means 'GIC v2' */
419 qemu_fdt_setprop_string(vms->fdt, "/intc", "compatible",
420 "arm,cortex-a15-gic");
421 qemu_fdt_setprop_sized_cells(vms->fdt, "/intc", "reg",
422 2, vms->memmap[VIRT_GIC_DIST].base,
423 2, vms->memmap[VIRT_GIC_DIST].size,
424 2, vms->memmap[VIRT_GIC_CPU].base,
425 2, vms->memmap[VIRT_GIC_CPU].size);
426 }
427
428 qemu_fdt_setprop_cell(vms->fdt, "/intc", "phandle", vms->gic_phandle);
429 }
430
431 static void fdt_add_pmu_nodes(const VirtMachineState *vms)
432 {
433 CPUState *cpu;
434 ARMCPU *armcpu;
435 uint32_t irqflags = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
436
437 CPU_FOREACH(cpu) {
438 armcpu = ARM_CPU(cpu);
439 if (!arm_feature(&armcpu->env, ARM_FEATURE_PMU)) {
440 return;
441 }
442 if (kvm_enabled()) {
443 if (kvm_irqchip_in_kernel()) {
444 kvm_arm_pmu_set_irq(cpu, PPI(VIRTUAL_PMU_IRQ));
445 }
446 kvm_arm_pmu_init(cpu);
447 }
448 }
449
450 if (vms->gic_version == 2) {
451 irqflags = deposit32(irqflags, GIC_FDT_IRQ_PPI_CPU_START,
452 GIC_FDT_IRQ_PPI_CPU_WIDTH,
453 (1 << vms->smp_cpus) - 1);
454 }
455
456 armcpu = ARM_CPU(qemu_get_cpu(0));
457 qemu_fdt_add_subnode(vms->fdt, "/pmu");
458 if (arm_feature(&armcpu->env, ARM_FEATURE_V8)) {
459 const char compat[] = "arm,armv8-pmuv3";
460 qemu_fdt_setprop(vms->fdt, "/pmu", "compatible",
461 compat, sizeof(compat));
462 qemu_fdt_setprop_cells(vms->fdt, "/pmu", "interrupts",
463 GIC_FDT_IRQ_TYPE_PPI, VIRTUAL_PMU_IRQ, irqflags);
464 }
465 }
466
467 static void create_its(VirtMachineState *vms, DeviceState *gicdev)
468 {
469 const char *itsclass = its_class_name();
470 DeviceState *dev;
471
472 if (!itsclass) {
473 /* Do nothing if not supported */
474 return;
475 }
476
477 dev = qdev_create(NULL, itsclass);
478
479 object_property_set_link(OBJECT(dev), OBJECT(gicdev), "parent-gicv3",
480 &error_abort);
481 qdev_init_nofail(dev);
482 sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_ITS].base);
483
484 fdt_add_its_gic_node(vms);
485 }
486
487 static void create_v2m(VirtMachineState *vms, qemu_irq *pic)
488 {
489 int i;
490 int irq = vms->irqmap[VIRT_GIC_V2M];
491 DeviceState *dev;
492
493 dev = qdev_create(NULL, "arm-gicv2m");
494 sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_V2M].base);
495 qdev_prop_set_uint32(dev, "base-spi", irq);
496 qdev_prop_set_uint32(dev, "num-spi", NUM_GICV2M_SPIS);
497 qdev_init_nofail(dev);
498
499 for (i = 0; i < NUM_GICV2M_SPIS; i++) {
500 sysbus_connect_irq(SYS_BUS_DEVICE(dev), i, pic[irq + i]);
501 }
502
503 fdt_add_v2m_gic_node(vms);
504 }
505
506 static void create_gic(VirtMachineState *vms, qemu_irq *pic)
507 {
508 /* We create a standalone GIC */
509 DeviceState *gicdev;
510 SysBusDevice *gicbusdev;
511 const char *gictype;
512 int type = vms->gic_version, i;
513
514 gictype = (type == 3) ? gicv3_class_name() : gic_class_name();
515
516 gicdev = qdev_create(NULL, gictype);
517 qdev_prop_set_uint32(gicdev, "revision", type);
518 qdev_prop_set_uint32(gicdev, "num-cpu", smp_cpus);
519 /* Note that the num-irq property counts both internal and external
520 * interrupts; there are always 32 of the former (mandated by GIC spec).
521 */
522 qdev_prop_set_uint32(gicdev, "num-irq", NUM_IRQS + 32);
523 if (!kvm_irqchip_in_kernel()) {
524 qdev_prop_set_bit(gicdev, "has-security-extensions", vms->secure);
525 }
526 qdev_init_nofail(gicdev);
527 gicbusdev = SYS_BUS_DEVICE(gicdev);
528 sysbus_mmio_map(gicbusdev, 0, vms->memmap[VIRT_GIC_DIST].base);
529 if (type == 3) {
530 sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_REDIST].base);
531 } else {
532 sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_CPU].base);
533 }
534
535 /* Wire the outputs from each CPU's generic timer and the GICv3
536 * maintenance interrupt signal to the appropriate GIC PPI inputs,
537 * and the GIC's IRQ/FIQ/VIRQ/VFIQ interrupt outputs to the CPU's inputs.
538 */
539 for (i = 0; i < smp_cpus; i++) {
540 DeviceState *cpudev = DEVICE(qemu_get_cpu(i));
541 int ppibase = NUM_IRQS + i * GIC_INTERNAL + GIC_NR_SGIS;
542 int irq;
543 /* Mapping from the output timer irq lines from the CPU to the
544 * GIC PPI inputs we use for the virt board.
545 */
546 const int timer_irq[] = {
547 [GTIMER_PHYS] = ARCH_TIMER_NS_EL1_IRQ,
548 [GTIMER_VIRT] = ARCH_TIMER_VIRT_IRQ,
549 [GTIMER_HYP] = ARCH_TIMER_NS_EL2_IRQ,
550 [GTIMER_SEC] = ARCH_TIMER_S_EL1_IRQ,
551 };
552
553 for (irq = 0; irq < ARRAY_SIZE(timer_irq); irq++) {
554 qdev_connect_gpio_out(cpudev, irq,
555 qdev_get_gpio_in(gicdev,
556 ppibase + timer_irq[irq]));
557 }
558
559 qdev_connect_gpio_out_named(cpudev, "gicv3-maintenance-interrupt", 0,
560 qdev_get_gpio_in(gicdev, ppibase
561 + ARCH_GICV3_MAINT_IRQ));
562 qdev_connect_gpio_out_named(cpudev, "pmu-interrupt", 0,
563 qdev_get_gpio_in(gicdev, ppibase
564 + VIRTUAL_PMU_IRQ));
565
566 sysbus_connect_irq(gicbusdev, i, qdev_get_gpio_in(cpudev, ARM_CPU_IRQ));
567 sysbus_connect_irq(gicbusdev, i + smp_cpus,
568 qdev_get_gpio_in(cpudev, ARM_CPU_FIQ));
569 sysbus_connect_irq(gicbusdev, i + 2 * smp_cpus,
570 qdev_get_gpio_in(cpudev, ARM_CPU_VIRQ));
571 sysbus_connect_irq(gicbusdev, i + 3 * smp_cpus,
572 qdev_get_gpio_in(cpudev, ARM_CPU_VFIQ));
573 }
574
575 for (i = 0; i < NUM_IRQS; i++) {
576 pic[i] = qdev_get_gpio_in(gicdev, i);
577 }
578
579 fdt_add_gic_node(vms);
580
581 if (type == 3 && vms->its) {
582 create_its(vms, gicdev);
583 } else if (type == 2) {
584 create_v2m(vms, pic);
585 }
586 }
587
588 static void create_uart(const VirtMachineState *vms, qemu_irq *pic, int uart,
589 MemoryRegion *mem, Chardev *chr)
590 {
591 char *nodename;
592 hwaddr base = vms->memmap[uart].base;
593 hwaddr size = vms->memmap[uart].size;
594 int irq = vms->irqmap[uart];
595 const char compat[] = "arm,pl011\0arm,primecell";
596 const char clocknames[] = "uartclk\0apb_pclk";
597 DeviceState *dev = qdev_create(NULL, "pl011");
598 SysBusDevice *s = SYS_BUS_DEVICE(dev);
599
600 qdev_prop_set_chr(dev, "chardev", chr);
601 qdev_init_nofail(dev);
602 memory_region_add_subregion(mem, base,
603 sysbus_mmio_get_region(s, 0));
604 sysbus_connect_irq(s, 0, pic[irq]);
605
606 nodename = g_strdup_printf("/pl011@%" PRIx64, base);
607 qemu_fdt_add_subnode(vms->fdt, nodename);
608 /* Note that we can't use setprop_string because of the embedded NUL */
609 qemu_fdt_setprop(vms->fdt, nodename, "compatible",
610 compat, sizeof(compat));
611 qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
612 2, base, 2, size);
613 qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupts",
614 GIC_FDT_IRQ_TYPE_SPI, irq,
615 GIC_FDT_IRQ_FLAGS_LEVEL_HI);
616 qemu_fdt_setprop_cells(vms->fdt, nodename, "clocks",
617 vms->clock_phandle, vms->clock_phandle);
618 qemu_fdt_setprop(vms->fdt, nodename, "clock-names",
619 clocknames, sizeof(clocknames));
620
621 if (uart == VIRT_UART) {
622 qemu_fdt_setprop_string(vms->fdt, "/chosen", "stdout-path", nodename);
623 } else {
624 /* Mark as not usable by the normal world */
625 qemu_fdt_setprop_string(vms->fdt, nodename, "status", "disabled");
626 qemu_fdt_setprop_string(vms->fdt, nodename, "secure-status", "okay");
627 }
628
629 g_free(nodename);
630 }
631
632 static void create_rtc(const VirtMachineState *vms, qemu_irq *pic)
633 {
634 char *nodename;
635 hwaddr base = vms->memmap[VIRT_RTC].base;
636 hwaddr size = vms->memmap[VIRT_RTC].size;
637 int irq = vms->irqmap[VIRT_RTC];
638 const char compat[] = "arm,pl031\0arm,primecell";
639
640 sysbus_create_simple("pl031", base, pic[irq]);
641
642 nodename = g_strdup_printf("/pl031@%" PRIx64, base);
643 qemu_fdt_add_subnode(vms->fdt, nodename);
644 qemu_fdt_setprop(vms->fdt, nodename, "compatible", compat, sizeof(compat));
645 qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
646 2, base, 2, size);
647 qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupts",
648 GIC_FDT_IRQ_TYPE_SPI, irq,
649 GIC_FDT_IRQ_FLAGS_LEVEL_HI);
650 qemu_fdt_setprop_cell(vms->fdt, nodename, "clocks", vms->clock_phandle);
651 qemu_fdt_setprop_string(vms->fdt, nodename, "clock-names", "apb_pclk");
652 g_free(nodename);
653 }
654
655 static DeviceState *gpio_key_dev;
656 static void virt_powerdown_req(Notifier *n, void *opaque)
657 {
658 /* use gpio Pin 3 for power button event */
659 qemu_set_irq(qdev_get_gpio_in(gpio_key_dev, 0), 1);
660 }
661
662 static Notifier virt_system_powerdown_notifier = {
663 .notify = virt_powerdown_req
664 };
665
666 static void create_gpio(const VirtMachineState *vms, qemu_irq *pic)
667 {
668 char *nodename;
669 DeviceState *pl061_dev;
670 hwaddr base = vms->memmap[VIRT_GPIO].base;
671 hwaddr size = vms->memmap[VIRT_GPIO].size;
672 int irq = vms->irqmap[VIRT_GPIO];
673 const char compat[] = "arm,pl061\0arm,primecell";
674
675 pl061_dev = sysbus_create_simple("pl061", base, pic[irq]);
676
677 uint32_t phandle = qemu_fdt_alloc_phandle(vms->fdt);
678 nodename = g_strdup_printf("/pl061@%" PRIx64, base);
679 qemu_fdt_add_subnode(vms->fdt, nodename);
680 qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
681 2, base, 2, size);
682 qemu_fdt_setprop(vms->fdt, nodename, "compatible", compat, sizeof(compat));
683 qemu_fdt_setprop_cell(vms->fdt, nodename, "#gpio-cells", 2);
684 qemu_fdt_setprop(vms->fdt, nodename, "gpio-controller", NULL, 0);
685 qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupts",
686 GIC_FDT_IRQ_TYPE_SPI, irq,
687 GIC_FDT_IRQ_FLAGS_LEVEL_HI);
688 qemu_fdt_setprop_cell(vms->fdt, nodename, "clocks", vms->clock_phandle);
689 qemu_fdt_setprop_string(vms->fdt, nodename, "clock-names", "apb_pclk");
690 qemu_fdt_setprop_cell(vms->fdt, nodename, "phandle", phandle);
691
692 gpio_key_dev = sysbus_create_simple("gpio-key", -1,
693 qdev_get_gpio_in(pl061_dev, 3));
694 qemu_fdt_add_subnode(vms->fdt, "/gpio-keys");
695 qemu_fdt_setprop_string(vms->fdt, "/gpio-keys", "compatible", "gpio-keys");
696 qemu_fdt_setprop_cell(vms->fdt, "/gpio-keys", "#size-cells", 0);
697 qemu_fdt_setprop_cell(vms->fdt, "/gpio-keys", "#address-cells", 1);
698
699 qemu_fdt_add_subnode(vms->fdt, "/gpio-keys/poweroff");
700 qemu_fdt_setprop_string(vms->fdt, "/gpio-keys/poweroff",
701 "label", "GPIO Key Poweroff");
702 qemu_fdt_setprop_cell(vms->fdt, "/gpio-keys/poweroff", "linux,code",
703 KEY_POWER);
704 qemu_fdt_setprop_cells(vms->fdt, "/gpio-keys/poweroff",
705 "gpios", phandle, 3, 0);
706
707 /* connect powerdown request */
708 qemu_register_powerdown_notifier(&virt_system_powerdown_notifier);
709
710 g_free(nodename);
711 }
712
713 static void create_virtio_devices(const VirtMachineState *vms, qemu_irq *pic)
714 {
715 int i;
716 hwaddr size = vms->memmap[VIRT_MMIO].size;
717
718 /* We create the transports in forwards order. Since qbus_realize()
719 * prepends (not appends) new child buses, the incrementing loop below will
720 * create a list of virtio-mmio buses with decreasing base addresses.
721 *
722 * When a -device option is processed from the command line,
723 * qbus_find_recursive() picks the next free virtio-mmio bus in forwards
724 * order. The upshot is that -device options in increasing command line
725 * order are mapped to virtio-mmio buses with decreasing base addresses.
726 *
727 * When this code was originally written, that arrangement ensured that the
728 * guest Linux kernel would give the lowest "name" (/dev/vda, eth0, etc) to
729 * the first -device on the command line. (The end-to-end order is a
730 * function of this loop, qbus_realize(), qbus_find_recursive(), and the
731 * guest kernel's name-to-address assignment strategy.)
732 *
733 * Meanwhile, the kernel's traversal seems to have been reversed; see eg.
734 * the message, if not necessarily the code, of commit 70161ff336.
735 * Therefore the loop now establishes the inverse of the original intent.
736 *
737 * Unfortunately, we can't counteract the kernel change by reversing the
738 * loop; it would break existing command lines.
739 *
740 * In any case, the kernel makes no guarantee about the stability of
741 * enumeration order of virtio devices (as demonstrated by it changing
742 * between kernel versions). For reliable and stable identification
743 * of disks users must use UUIDs or similar mechanisms.
744 */
745 for (i = 0; i < NUM_VIRTIO_TRANSPORTS; i++) {
746 int irq = vms->irqmap[VIRT_MMIO] + i;
747 hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
748
749 sysbus_create_simple("virtio-mmio", base, pic[irq]);
750 }
751
752 /* We add dtb nodes in reverse order so that they appear in the finished
753 * device tree lowest address first.
754 *
755 * Note that this mapping is independent of the loop above. The previous
756 * loop influences virtio device to virtio transport assignment, whereas
757 * this loop controls how virtio transports are laid out in the dtb.
758 */
759 for (i = NUM_VIRTIO_TRANSPORTS - 1; i >= 0; i--) {
760 char *nodename;
761 int irq = vms->irqmap[VIRT_MMIO] + i;
762 hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
763
764 nodename = g_strdup_printf("/virtio_mmio@%" PRIx64, base);
765 qemu_fdt_add_subnode(vms->fdt, nodename);
766 qemu_fdt_setprop_string(vms->fdt, nodename,
767 "compatible", "virtio,mmio");
768 qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
769 2, base, 2, size);
770 qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupts",
771 GIC_FDT_IRQ_TYPE_SPI, irq,
772 GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
773 qemu_fdt_setprop(vms->fdt, nodename, "dma-coherent", NULL, 0);
774 g_free(nodename);
775 }
776 }
777
778 static void create_one_flash(const char *name, hwaddr flashbase,
779 hwaddr flashsize, const char *file,
780 MemoryRegion *sysmem)
781 {
782 /* Create and map a single flash device. We use the same
783 * parameters as the flash devices on the Versatile Express board.
784 */
785 DriveInfo *dinfo = drive_get_next(IF_PFLASH);
786 DeviceState *dev = qdev_create(NULL, "cfi.pflash01");
787 SysBusDevice *sbd = SYS_BUS_DEVICE(dev);
788 const uint64_t sectorlength = 256 * 1024;
789
790 if (dinfo) {
791 qdev_prop_set_drive(dev, "drive", blk_by_legacy_dinfo(dinfo),
792 &error_abort);
793 }
794
795 qdev_prop_set_uint32(dev, "num-blocks", flashsize / sectorlength);
796 qdev_prop_set_uint64(dev, "sector-length", sectorlength);
797 qdev_prop_set_uint8(dev, "width", 4);
798 qdev_prop_set_uint8(dev, "device-width", 2);
799 qdev_prop_set_bit(dev, "big-endian", false);
800 qdev_prop_set_uint16(dev, "id0", 0x89);
801 qdev_prop_set_uint16(dev, "id1", 0x18);
802 qdev_prop_set_uint16(dev, "id2", 0x00);
803 qdev_prop_set_uint16(dev, "id3", 0x00);
804 qdev_prop_set_string(dev, "name", name);
805 qdev_init_nofail(dev);
806
807 memory_region_add_subregion(sysmem, flashbase,
808 sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 0));
809
810 if (file) {
811 char *fn;
812 int image_size;
813
814 if (drive_get(IF_PFLASH, 0, 0)) {
815 error_report("The contents of the first flash device may be "
816 "specified with -bios or with -drive if=pflash... "
817 "but you cannot use both options at once");
818 exit(1);
819 }
820 fn = qemu_find_file(QEMU_FILE_TYPE_BIOS, file);
821 if (!fn) {
822 error_report("Could not find ROM image '%s'", file);
823 exit(1);
824 }
825 image_size = load_image_mr(fn, sysbus_mmio_get_region(sbd, 0));
826 g_free(fn);
827 if (image_size < 0) {
828 error_report("Could not load ROM image '%s'", file);
829 exit(1);
830 }
831 }
832 }
833
834 static void create_flash(const VirtMachineState *vms,
835 MemoryRegion *sysmem,
836 MemoryRegion *secure_sysmem)
837 {
838 /* Create two flash devices to fill the VIRT_FLASH space in the memmap.
839 * Any file passed via -bios goes in the first of these.
840 * sysmem is the system memory space. secure_sysmem is the secure view
841 * of the system, and the first flash device should be made visible only
842 * there. The second flash device is visible to both secure and nonsecure.
843 * If sysmem == secure_sysmem this means there is no separate Secure
844 * address space and both flash devices are generally visible.
845 */
846 hwaddr flashsize = vms->memmap[VIRT_FLASH].size / 2;
847 hwaddr flashbase = vms->memmap[VIRT_FLASH].base;
848 char *nodename;
849
850 create_one_flash("virt.flash0", flashbase, flashsize,
851 bios_name, secure_sysmem);
852 create_one_flash("virt.flash1", flashbase + flashsize, flashsize,
853 NULL, sysmem);
854
855 if (sysmem == secure_sysmem) {
856 /* Report both flash devices as a single node in the DT */
857 nodename = g_strdup_printf("/flash@%" PRIx64, flashbase);
858 qemu_fdt_add_subnode(vms->fdt, nodename);
859 qemu_fdt_setprop_string(vms->fdt, nodename, "compatible", "cfi-flash");
860 qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
861 2, flashbase, 2, flashsize,
862 2, flashbase + flashsize, 2, flashsize);
863 qemu_fdt_setprop_cell(vms->fdt, nodename, "bank-width", 4);
864 g_free(nodename);
865 } else {
866 /* Report the devices as separate nodes so we can mark one as
867 * only visible to the secure world.
868 */
869 nodename = g_strdup_printf("/secflash@%" PRIx64, flashbase);
870 qemu_fdt_add_subnode(vms->fdt, nodename);
871 qemu_fdt_setprop_string(vms->fdt, nodename, "compatible", "cfi-flash");
872 qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
873 2, flashbase, 2, flashsize);
874 qemu_fdt_setprop_cell(vms->fdt, nodename, "bank-width", 4);
875 qemu_fdt_setprop_string(vms->fdt, nodename, "status", "disabled");
876 qemu_fdt_setprop_string(vms->fdt, nodename, "secure-status", "okay");
877 g_free(nodename);
878
879 nodename = g_strdup_printf("/flash@%" PRIx64, flashbase);
880 qemu_fdt_add_subnode(vms->fdt, nodename);
881 qemu_fdt_setprop_string(vms->fdt, nodename, "compatible", "cfi-flash");
882 qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
883 2, flashbase + flashsize, 2, flashsize);
884 qemu_fdt_setprop_cell(vms->fdt, nodename, "bank-width", 4);
885 g_free(nodename);
886 }
887 }
888
889 static FWCfgState *create_fw_cfg(const VirtMachineState *vms, AddressSpace *as)
890 {
891 hwaddr base = vms->memmap[VIRT_FW_CFG].base;
892 hwaddr size = vms->memmap[VIRT_FW_CFG].size;
893 FWCfgState *fw_cfg;
894 char *nodename;
895
896 fw_cfg = fw_cfg_init_mem_wide(base + 8, base, 8, base + 16, as);
897 fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, (uint16_t)smp_cpus);
898
899 nodename = g_strdup_printf("/fw-cfg@%" PRIx64, base);
900 qemu_fdt_add_subnode(vms->fdt, nodename);
901 qemu_fdt_setprop_string(vms->fdt, nodename,
902 "compatible", "qemu,fw-cfg-mmio");
903 qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
904 2, base, 2, size);
905 qemu_fdt_setprop(vms->fdt, nodename, "dma-coherent", NULL, 0);
906 g_free(nodename);
907 return fw_cfg;
908 }
909
910 static void create_pcie_irq_map(const VirtMachineState *vms,
911 uint32_t gic_phandle,
912 int first_irq, const char *nodename)
913 {
914 int devfn, pin;
915 uint32_t full_irq_map[4 * 4 * 10] = { 0 };
916 uint32_t *irq_map = full_irq_map;
917
918 for (devfn = 0; devfn <= 0x18; devfn += 0x8) {
919 for (pin = 0; pin < 4; pin++) {
920 int irq_type = GIC_FDT_IRQ_TYPE_SPI;
921 int irq_nr = first_irq + ((pin + PCI_SLOT(devfn)) % PCI_NUM_PINS);
922 int irq_level = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
923 int i;
924
925 uint32_t map[] = {
926 devfn << 8, 0, 0, /* devfn */
927 pin + 1, /* PCI pin */
928 gic_phandle, 0, 0, irq_type, irq_nr, irq_level }; /* GIC irq */
929
930 /* Convert map to big endian */
931 for (i = 0; i < 10; i++) {
932 irq_map[i] = cpu_to_be32(map[i]);
933 }
934 irq_map += 10;
935 }
936 }
937
938 qemu_fdt_setprop(vms->fdt, nodename, "interrupt-map",
939 full_irq_map, sizeof(full_irq_map));
940
941 qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupt-map-mask",
942 0x1800, 0, 0, /* devfn (PCI_SLOT(3)) */
943 0x7 /* PCI irq */);
944 }
945
946 static void create_smmu(const VirtMachineState *vms, qemu_irq *pic,
947 PCIBus *bus)
948 {
949 char *node;
950 const char compat[] = "arm,smmu-v3";
951 int irq = vms->irqmap[VIRT_SMMU];
952 int i;
953 hwaddr base = vms->memmap[VIRT_SMMU].base;
954 hwaddr size = vms->memmap[VIRT_SMMU].size;
955 const char irq_names[] = "eventq\0priq\0cmdq-sync\0gerror";
956 DeviceState *dev;
957
958 if (vms->iommu != VIRT_IOMMU_SMMUV3 || !vms->iommu_phandle) {
959 return;
960 }
961
962 dev = qdev_create(NULL, "arm-smmuv3");
963
964 object_property_set_link(OBJECT(dev), OBJECT(bus), "primary-bus",
965 &error_abort);
966 qdev_init_nofail(dev);
967 sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, base);
968 for (i = 0; i < NUM_SMMU_IRQS; i++) {
969 sysbus_connect_irq(SYS_BUS_DEVICE(dev), i, pic[irq + i]);
970 }
971
972 node = g_strdup_printf("/smmuv3@%" PRIx64, base);
973 qemu_fdt_add_subnode(vms->fdt, node);
974 qemu_fdt_setprop(vms->fdt, node, "compatible", compat, sizeof(compat));
975 qemu_fdt_setprop_sized_cells(vms->fdt, node, "reg", 2, base, 2, size);
976
977 qemu_fdt_setprop_cells(vms->fdt, node, "interrupts",
978 GIC_FDT_IRQ_TYPE_SPI, irq , GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
979 GIC_FDT_IRQ_TYPE_SPI, irq + 1, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
980 GIC_FDT_IRQ_TYPE_SPI, irq + 2, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
981 GIC_FDT_IRQ_TYPE_SPI, irq + 3, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
982
983 qemu_fdt_setprop(vms->fdt, node, "interrupt-names", irq_names,
984 sizeof(irq_names));
985
986 qemu_fdt_setprop_cell(vms->fdt, node, "clocks", vms->clock_phandle);
987 qemu_fdt_setprop_string(vms->fdt, node, "clock-names", "apb_pclk");
988 qemu_fdt_setprop(vms->fdt, node, "dma-coherent", NULL, 0);
989
990 qemu_fdt_setprop_cell(vms->fdt, node, "#iommu-cells", 1);
991
992 qemu_fdt_setprop_cell(vms->fdt, node, "phandle", vms->iommu_phandle);
993 g_free(node);
994 }
995
996 static void create_pcie(VirtMachineState *vms, qemu_irq *pic)
997 {
998 hwaddr base_mmio = vms->memmap[VIRT_PCIE_MMIO].base;
999 hwaddr size_mmio = vms->memmap[VIRT_PCIE_MMIO].size;
1000 hwaddr base_mmio_high = vms->memmap[VIRT_PCIE_MMIO_HIGH].base;
1001 hwaddr size_mmio_high = vms->memmap[VIRT_PCIE_MMIO_HIGH].size;
1002 hwaddr base_pio = vms->memmap[VIRT_PCIE_PIO].base;
1003 hwaddr size_pio = vms->memmap[VIRT_PCIE_PIO].size;
1004 hwaddr base_ecam = vms->memmap[VIRT_PCIE_ECAM].base;
1005 hwaddr size_ecam = vms->memmap[VIRT_PCIE_ECAM].size;
1006 hwaddr base = base_mmio;
1007 int nr_pcie_buses = size_ecam / PCIE_MMCFG_SIZE_MIN;
1008 int irq = vms->irqmap[VIRT_PCIE];
1009 MemoryRegion *mmio_alias;
1010 MemoryRegion *mmio_reg;
1011 MemoryRegion *ecam_alias;
1012 MemoryRegion *ecam_reg;
1013 DeviceState *dev;
1014 char *nodename;
1015 int i;
1016 PCIHostState *pci;
1017
1018 dev = qdev_create(NULL, TYPE_GPEX_HOST);
1019 qdev_init_nofail(dev);
1020
1021 /* Map only the first size_ecam bytes of ECAM space */
1022 ecam_alias = g_new0(MemoryRegion, 1);
1023 ecam_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 0);
1024 memory_region_init_alias(ecam_alias, OBJECT(dev), "pcie-ecam",
1025 ecam_reg, 0, size_ecam);
1026 memory_region_add_subregion(get_system_memory(), base_ecam, ecam_alias);
1027
1028 /* Map the MMIO window into system address space so as to expose
1029 * the section of PCI MMIO space which starts at the same base address
1030 * (ie 1:1 mapping for that part of PCI MMIO space visible through
1031 * the window).
1032 */
1033 mmio_alias = g_new0(MemoryRegion, 1);
1034 mmio_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 1);
1035 memory_region_init_alias(mmio_alias, OBJECT(dev), "pcie-mmio",
1036 mmio_reg, base_mmio, size_mmio);
1037 memory_region_add_subregion(get_system_memory(), base_mmio, mmio_alias);
1038
1039 if (vms->highmem) {
1040 /* Map high MMIO space */
1041 MemoryRegion *high_mmio_alias = g_new0(MemoryRegion, 1);
1042
1043 memory_region_init_alias(high_mmio_alias, OBJECT(dev), "pcie-mmio-high",
1044 mmio_reg, base_mmio_high, size_mmio_high);
1045 memory_region_add_subregion(get_system_memory(), base_mmio_high,
1046 high_mmio_alias);
1047 }
1048
1049 /* Map IO port space */
1050 sysbus_mmio_map(SYS_BUS_DEVICE(dev), 2, base_pio);
1051
1052 for (i = 0; i < GPEX_NUM_IRQS; i++) {
1053 sysbus_connect_irq(SYS_BUS_DEVICE(dev), i, pic[irq + i]);
1054 gpex_set_irq_num(GPEX_HOST(dev), i, irq + i);
1055 }
1056
1057 pci = PCI_HOST_BRIDGE(dev);
1058 if (pci->bus) {
1059 for (i = 0; i < nb_nics; i++) {
1060 NICInfo *nd = &nd_table[i];
1061
1062 if (!nd->model) {
1063 nd->model = g_strdup("virtio");
1064 }
1065
1066 pci_nic_init_nofail(nd, pci->bus, nd->model, NULL);
1067 }
1068 }
1069
1070 nodename = g_strdup_printf("/pcie@%" PRIx64, base);
1071 qemu_fdt_add_subnode(vms->fdt, nodename);
1072 qemu_fdt_setprop_string(vms->fdt, nodename,
1073 "compatible", "pci-host-ecam-generic");
1074 qemu_fdt_setprop_string(vms->fdt, nodename, "device_type", "pci");
1075 qemu_fdt_setprop_cell(vms->fdt, nodename, "#address-cells", 3);
1076 qemu_fdt_setprop_cell(vms->fdt, nodename, "#size-cells", 2);
1077 qemu_fdt_setprop_cell(vms->fdt, nodename, "linux,pci-domain", 0);
1078 qemu_fdt_setprop_cells(vms->fdt, nodename, "bus-range", 0,
1079 nr_pcie_buses - 1);
1080 qemu_fdt_setprop(vms->fdt, nodename, "dma-coherent", NULL, 0);
1081
1082 if (vms->msi_phandle) {
1083 qemu_fdt_setprop_cells(vms->fdt, nodename, "msi-parent",
1084 vms->msi_phandle);
1085 }
1086
1087 qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
1088 2, base_ecam, 2, size_ecam);
1089
1090 if (vms->highmem) {
1091 qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "ranges",
1092 1, FDT_PCI_RANGE_IOPORT, 2, 0,
1093 2, base_pio, 2, size_pio,
1094 1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1095 2, base_mmio, 2, size_mmio,
1096 1, FDT_PCI_RANGE_MMIO_64BIT,
1097 2, base_mmio_high,
1098 2, base_mmio_high, 2, size_mmio_high);
1099 } else {
1100 qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "ranges",
1101 1, FDT_PCI_RANGE_IOPORT, 2, 0,
1102 2, base_pio, 2, size_pio,
1103 1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1104 2, base_mmio, 2, size_mmio);
1105 }
1106
1107 qemu_fdt_setprop_cell(vms->fdt, nodename, "#interrupt-cells", 1);
1108 create_pcie_irq_map(vms, vms->gic_phandle, irq, nodename);
1109
1110 if (vms->iommu) {
1111 vms->iommu_phandle = qemu_fdt_alloc_phandle(vms->fdt);
1112
1113 create_smmu(vms, pic, pci->bus);
1114
1115 qemu_fdt_setprop_cells(vms->fdt, nodename, "iommu-map",
1116 0x0, vms->iommu_phandle, 0x0, 0x10000);
1117 }
1118
1119 g_free(nodename);
1120 }
1121
1122 static void create_platform_bus(VirtMachineState *vms, qemu_irq *pic)
1123 {
1124 DeviceState *dev;
1125 SysBusDevice *s;
1126 int i;
1127 MemoryRegion *sysmem = get_system_memory();
1128
1129 dev = qdev_create(NULL, TYPE_PLATFORM_BUS_DEVICE);
1130 dev->id = TYPE_PLATFORM_BUS_DEVICE;
1131 qdev_prop_set_uint32(dev, "num_irqs", PLATFORM_BUS_NUM_IRQS);
1132 qdev_prop_set_uint32(dev, "mmio_size", vms->memmap[VIRT_PLATFORM_BUS].size);
1133 qdev_init_nofail(dev);
1134 vms->platform_bus_dev = dev;
1135
1136 s = SYS_BUS_DEVICE(dev);
1137 for (i = 0; i < PLATFORM_BUS_NUM_IRQS; i++) {
1138 int irqn = vms->irqmap[VIRT_PLATFORM_BUS] + i;
1139 sysbus_connect_irq(s, i, pic[irqn]);
1140 }
1141
1142 memory_region_add_subregion(sysmem,
1143 vms->memmap[VIRT_PLATFORM_BUS].base,
1144 sysbus_mmio_get_region(s, 0));
1145 }
1146
1147 static void create_secure_ram(VirtMachineState *vms,
1148 MemoryRegion *secure_sysmem)
1149 {
1150 MemoryRegion *secram = g_new(MemoryRegion, 1);
1151 char *nodename;
1152 hwaddr base = vms->memmap[VIRT_SECURE_MEM].base;
1153 hwaddr size = vms->memmap[VIRT_SECURE_MEM].size;
1154
1155 memory_region_init_ram(secram, NULL, "virt.secure-ram", size,
1156 &error_fatal);
1157 memory_region_add_subregion(secure_sysmem, base, secram);
1158
1159 nodename = g_strdup_printf("/secram@%" PRIx64, base);
1160 qemu_fdt_add_subnode(vms->fdt, nodename);
1161 qemu_fdt_setprop_string(vms->fdt, nodename, "device_type", "memory");
1162 qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg", 2, base, 2, size);
1163 qemu_fdt_setprop_string(vms->fdt, nodename, "status", "disabled");
1164 qemu_fdt_setprop_string(vms->fdt, nodename, "secure-status", "okay");
1165
1166 g_free(nodename);
1167 }
1168
1169 static void *machvirt_dtb(const struct arm_boot_info *binfo, int *fdt_size)
1170 {
1171 const VirtMachineState *board = container_of(binfo, VirtMachineState,
1172 bootinfo);
1173
1174 *fdt_size = board->fdt_size;
1175 return board->fdt;
1176 }
1177
1178 static void virt_build_smbios(VirtMachineState *vms)
1179 {
1180 MachineClass *mc = MACHINE_GET_CLASS(vms);
1181 VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1182 uint8_t *smbios_tables, *smbios_anchor;
1183 size_t smbios_tables_len, smbios_anchor_len;
1184 const char *product = "QEMU Virtual Machine";
1185
1186 if (!vms->fw_cfg) {
1187 return;
1188 }
1189
1190 if (kvm_enabled()) {
1191 product = "KVM Virtual Machine";
1192 }
1193
1194 smbios_set_defaults("QEMU", product,
1195 vmc->smbios_old_sys_ver ? "1.0" : mc->name, false,
1196 true, SMBIOS_ENTRY_POINT_30);
1197
1198 smbios_get_tables(NULL, 0, &smbios_tables, &smbios_tables_len,
1199 &smbios_anchor, &smbios_anchor_len);
1200
1201 if (smbios_anchor) {
1202 fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-tables",
1203 smbios_tables, smbios_tables_len);
1204 fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-anchor",
1205 smbios_anchor, smbios_anchor_len);
1206 }
1207 }
1208
1209 static
1210 void virt_machine_done(Notifier *notifier, void *data)
1211 {
1212 VirtMachineState *vms = container_of(notifier, VirtMachineState,
1213 machine_done);
1214 ARMCPU *cpu = ARM_CPU(first_cpu);
1215 struct arm_boot_info *info = &vms->bootinfo;
1216 AddressSpace *as = arm_boot_address_space(cpu, info);
1217
1218 /*
1219 * If the user provided a dtb, we assume the dynamic sysbus nodes
1220 * already are integrated there. This corresponds to a use case where
1221 * the dynamic sysbus nodes are complex and their generation is not yet
1222 * supported. In that case the user can take charge of the guest dt
1223 * while qemu takes charge of the qom stuff.
1224 */
1225 if (info->dtb_filename == NULL) {
1226 platform_bus_add_all_fdt_nodes(vms->fdt, "/intc",
1227 vms->memmap[VIRT_PLATFORM_BUS].base,
1228 vms->memmap[VIRT_PLATFORM_BUS].size,
1229 vms->irqmap[VIRT_PLATFORM_BUS]);
1230 }
1231 if (arm_load_dtb(info->dtb_start, info, info->dtb_limit, as) < 0) {
1232 exit(1);
1233 }
1234
1235 virt_acpi_setup(vms);
1236 virt_build_smbios(vms);
1237 }
1238
1239 static uint64_t virt_cpu_mp_affinity(VirtMachineState *vms, int idx)
1240 {
1241 uint8_t clustersz = ARM_DEFAULT_CPUS_PER_CLUSTER;
1242 VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1243
1244 if (!vmc->disallow_affinity_adjustment) {
1245 /* Adjust MPIDR like 64-bit KVM hosts, which incorporate the
1246 * GIC's target-list limitations. 32-bit KVM hosts currently
1247 * always create clusters of 4 CPUs, but that is expected to
1248 * change when they gain support for gicv3. When KVM is enabled
1249 * it will override the changes we make here, therefore our
1250 * purposes are to make TCG consistent (with 64-bit KVM hosts)
1251 * and to improve SGI efficiency.
1252 */
1253 if (vms->gic_version == 3) {
1254 clustersz = GICV3_TARGETLIST_BITS;
1255 } else {
1256 clustersz = GIC_TARGETLIST_BITS;
1257 }
1258 }
1259 return arm_cpu_mp_affinity(idx, clustersz);
1260 }
1261
1262 static void machvirt_init(MachineState *machine)
1263 {
1264 VirtMachineState *vms = VIRT_MACHINE(machine);
1265 VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(machine);
1266 MachineClass *mc = MACHINE_GET_CLASS(machine);
1267 const CPUArchIdList *possible_cpus;
1268 qemu_irq pic[NUM_IRQS];
1269 MemoryRegion *sysmem = get_system_memory();
1270 MemoryRegion *secure_sysmem = NULL;
1271 int n, virt_max_cpus;
1272 MemoryRegion *ram = g_new(MemoryRegion, 1);
1273 bool firmware_loaded = bios_name || drive_get(IF_PFLASH, 0, 0);
1274
1275 /* We can probe only here because during property set
1276 * KVM is not available yet
1277 */
1278 if (vms->gic_version <= 0) {
1279 /* "host" or "max" */
1280 if (!kvm_enabled()) {
1281 if (vms->gic_version == 0) {
1282 error_report("gic-version=host requires KVM");
1283 exit(1);
1284 } else {
1285 /* "max": currently means 3 for TCG */
1286 vms->gic_version = 3;
1287 }
1288 } else {
1289 vms->gic_version = kvm_arm_vgic_probe();
1290 if (!vms->gic_version) {
1291 error_report(
1292 "Unable to determine GIC version supported by host");
1293 exit(1);
1294 }
1295 }
1296 }
1297
1298 if (!cpu_type_valid(machine->cpu_type)) {
1299 error_report("mach-virt: CPU type %s not supported", machine->cpu_type);
1300 exit(1);
1301 }
1302
1303 /* If we have an EL3 boot ROM then the assumption is that it will
1304 * implement PSCI itself, so disable QEMU's internal implementation
1305 * so it doesn't get in the way. Instead of starting secondary
1306 * CPUs in PSCI powerdown state we will start them all running and
1307 * let the boot ROM sort them out.
1308 * The usual case is that we do use QEMU's PSCI implementation;
1309 * if the guest has EL2 then we will use SMC as the conduit,
1310 * and otherwise we will use HVC (for backwards compatibility and
1311 * because if we're using KVM then we must use HVC).
1312 */
1313 if (vms->secure && firmware_loaded) {
1314 vms->psci_conduit = QEMU_PSCI_CONDUIT_DISABLED;
1315 } else if (vms->virt) {
1316 vms->psci_conduit = QEMU_PSCI_CONDUIT_SMC;
1317 } else {
1318 vms->psci_conduit = QEMU_PSCI_CONDUIT_HVC;
1319 }
1320
1321 /* The maximum number of CPUs depends on the GIC version, or on how
1322 * many redistributors we can fit into the memory map.
1323 */
1324 if (vms->gic_version == 3) {
1325 virt_max_cpus = vms->memmap[VIRT_GIC_REDIST].size / 0x20000;
1326 } else {
1327 virt_max_cpus = GIC_NCPU;
1328 }
1329
1330 if (max_cpus > virt_max_cpus) {
1331 error_report("Number of SMP CPUs requested (%d) exceeds max CPUs "
1332 "supported by machine 'mach-virt' (%d)",
1333 max_cpus, virt_max_cpus);
1334 exit(1);
1335 }
1336
1337 vms->smp_cpus = smp_cpus;
1338
1339 if (machine->ram_size > vms->memmap[VIRT_MEM].size) {
1340 error_report("mach-virt: cannot model more than %dGB RAM", RAMLIMIT_GB);
1341 exit(1);
1342 }
1343
1344 if (vms->virt && kvm_enabled()) {
1345 error_report("mach-virt: KVM does not support providing "
1346 "Virtualization extensions to the guest CPU");
1347 exit(1);
1348 }
1349
1350 if (vms->secure) {
1351 if (kvm_enabled()) {
1352 error_report("mach-virt: KVM does not support Security extensions");
1353 exit(1);
1354 }
1355
1356 /* The Secure view of the world is the same as the NonSecure,
1357 * but with a few extra devices. Create it as a container region
1358 * containing the system memory at low priority; any secure-only
1359 * devices go in at higher priority and take precedence.
1360 */
1361 secure_sysmem = g_new(MemoryRegion, 1);
1362 memory_region_init(secure_sysmem, OBJECT(machine), "secure-memory",
1363 UINT64_MAX);
1364 memory_region_add_subregion_overlap(secure_sysmem, 0, sysmem, -1);
1365 }
1366
1367 create_fdt(vms);
1368
1369 possible_cpus = mc->possible_cpu_arch_ids(machine);
1370 for (n = 0; n < possible_cpus->len; n++) {
1371 Object *cpuobj;
1372 CPUState *cs;
1373
1374 if (n >= smp_cpus) {
1375 break;
1376 }
1377
1378 cpuobj = object_new(possible_cpus->cpus[n].type);
1379 object_property_set_int(cpuobj, possible_cpus->cpus[n].arch_id,
1380 "mp-affinity", NULL);
1381
1382 cs = CPU(cpuobj);
1383 cs->cpu_index = n;
1384
1385 numa_cpu_pre_plug(&possible_cpus->cpus[cs->cpu_index], DEVICE(cpuobj),
1386 &error_fatal);
1387
1388 if (!vms->secure) {
1389 object_property_set_bool(cpuobj, false, "has_el3", NULL);
1390 }
1391
1392 if (!vms->virt && object_property_find(cpuobj, "has_el2", NULL)) {
1393 object_property_set_bool(cpuobj, false, "has_el2", NULL);
1394 }
1395
1396 if (vms->psci_conduit != QEMU_PSCI_CONDUIT_DISABLED) {
1397 object_property_set_int(cpuobj, vms->psci_conduit,
1398 "psci-conduit", NULL);
1399
1400 /* Secondary CPUs start in PSCI powered-down state */
1401 if (n > 0) {
1402 object_property_set_bool(cpuobj, true,
1403 "start-powered-off", NULL);
1404 }
1405 }
1406
1407 if (vmc->no_pmu && object_property_find(cpuobj, "pmu", NULL)) {
1408 object_property_set_bool(cpuobj, false, "pmu", NULL);
1409 }
1410
1411 if (object_property_find(cpuobj, "reset-cbar", NULL)) {
1412 object_property_set_int(cpuobj, vms->memmap[VIRT_CPUPERIPHS].base,
1413 "reset-cbar", &error_abort);
1414 }
1415
1416 object_property_set_link(cpuobj, OBJECT(sysmem), "memory",
1417 &error_abort);
1418 if (vms->secure) {
1419 object_property_set_link(cpuobj, OBJECT(secure_sysmem),
1420 "secure-memory", &error_abort);
1421 }
1422
1423 object_property_set_bool(cpuobj, true, "realized", &error_fatal);
1424 object_unref(cpuobj);
1425 }
1426 fdt_add_timer_nodes(vms);
1427 fdt_add_cpu_nodes(vms);
1428
1429 memory_region_allocate_system_memory(ram, NULL, "mach-virt.ram",
1430 machine->ram_size);
1431 memory_region_add_subregion(sysmem, vms->memmap[VIRT_MEM].base, ram);
1432
1433 create_flash(vms, sysmem, secure_sysmem ? secure_sysmem : sysmem);
1434
1435 create_gic(vms, pic);
1436
1437 fdt_add_pmu_nodes(vms);
1438
1439 create_uart(vms, pic, VIRT_UART, sysmem, serial_hd(0));
1440
1441 if (vms->secure) {
1442 create_secure_ram(vms, secure_sysmem);
1443 create_uart(vms, pic, VIRT_SECURE_UART, secure_sysmem, serial_hd(1));
1444 }
1445
1446 create_rtc(vms, pic);
1447
1448 create_pcie(vms, pic);
1449
1450 create_gpio(vms, pic);
1451
1452 /* Create mmio transports, so the user can create virtio backends
1453 * (which will be automatically plugged in to the transports). If
1454 * no backend is created the transport will just sit harmlessly idle.
1455 */
1456 create_virtio_devices(vms, pic);
1457
1458 vms->fw_cfg = create_fw_cfg(vms, &address_space_memory);
1459 rom_set_fw(vms->fw_cfg);
1460
1461 create_platform_bus(vms, pic);
1462
1463 vms->bootinfo.ram_size = machine->ram_size;
1464 vms->bootinfo.kernel_filename = machine->kernel_filename;
1465 vms->bootinfo.kernel_cmdline = machine->kernel_cmdline;
1466 vms->bootinfo.initrd_filename = machine->initrd_filename;
1467 vms->bootinfo.nb_cpus = smp_cpus;
1468 vms->bootinfo.board_id = -1;
1469 vms->bootinfo.loader_start = vms->memmap[VIRT_MEM].base;
1470 vms->bootinfo.get_dtb = machvirt_dtb;
1471 vms->bootinfo.skip_dtb_autoload = true;
1472 vms->bootinfo.firmware_loaded = firmware_loaded;
1473 arm_load_kernel(ARM_CPU(first_cpu), &vms->bootinfo);
1474
1475 vms->machine_done.notify = virt_machine_done;
1476 qemu_add_machine_init_done_notifier(&vms->machine_done);
1477 }
1478
1479 static bool virt_get_secure(Object *obj, Error **errp)
1480 {
1481 VirtMachineState *vms = VIRT_MACHINE(obj);
1482
1483 return vms->secure;
1484 }
1485
1486 static void virt_set_secure(Object *obj, bool value, Error **errp)
1487 {
1488 VirtMachineState *vms = VIRT_MACHINE(obj);
1489
1490 vms->secure = value;
1491 }
1492
1493 static bool virt_get_virt(Object *obj, Error **errp)
1494 {
1495 VirtMachineState *vms = VIRT_MACHINE(obj);
1496
1497 return vms->virt;
1498 }
1499
1500 static void virt_set_virt(Object *obj, bool value, Error **errp)
1501 {
1502 VirtMachineState *vms = VIRT_MACHINE(obj);
1503
1504 vms->virt = value;
1505 }
1506
1507 static bool virt_get_highmem(Object *obj, Error **errp)
1508 {
1509 VirtMachineState *vms = VIRT_MACHINE(obj);
1510
1511 return vms->highmem;
1512 }
1513
1514 static void virt_set_highmem(Object *obj, bool value, Error **errp)
1515 {
1516 VirtMachineState *vms = VIRT_MACHINE(obj);
1517
1518 vms->highmem = value;
1519 }
1520
1521 static bool virt_get_its(Object *obj, Error **errp)
1522 {
1523 VirtMachineState *vms = VIRT_MACHINE(obj);
1524
1525 return vms->its;
1526 }
1527
1528 static void virt_set_its(Object *obj, bool value, Error **errp)
1529 {
1530 VirtMachineState *vms = VIRT_MACHINE(obj);
1531
1532 vms->its = value;
1533 }
1534
1535 static char *virt_get_gic_version(Object *obj, Error **errp)
1536 {
1537 VirtMachineState *vms = VIRT_MACHINE(obj);
1538 const char *val = vms->gic_version == 3 ? "3" : "2";
1539
1540 return g_strdup(val);
1541 }
1542
1543 static void virt_set_gic_version(Object *obj, const char *value, Error **errp)
1544 {
1545 VirtMachineState *vms = VIRT_MACHINE(obj);
1546
1547 if (!strcmp(value, "3")) {
1548 vms->gic_version = 3;
1549 } else if (!strcmp(value, "2")) {
1550 vms->gic_version = 2;
1551 } else if (!strcmp(value, "host")) {
1552 vms->gic_version = 0; /* Will probe later */
1553 } else if (!strcmp(value, "max")) {
1554 vms->gic_version = -1; /* Will probe later */
1555 } else {
1556 error_setg(errp, "Invalid gic-version value");
1557 error_append_hint(errp, "Valid values are 3, 2, host, max.\n");
1558 }
1559 }
1560
1561 static char *virt_get_iommu(Object *obj, Error **errp)
1562 {
1563 VirtMachineState *vms = VIRT_MACHINE(obj);
1564
1565 switch (vms->iommu) {
1566 case VIRT_IOMMU_NONE:
1567 return g_strdup("none");
1568 case VIRT_IOMMU_SMMUV3:
1569 return g_strdup("smmuv3");
1570 default:
1571 g_assert_not_reached();
1572 }
1573 }
1574
1575 static void virt_set_iommu(Object *obj, const char *value, Error **errp)
1576 {
1577 VirtMachineState *vms = VIRT_MACHINE(obj);
1578
1579 if (!strcmp(value, "smmuv3")) {
1580 vms->iommu = VIRT_IOMMU_SMMUV3;
1581 } else if (!strcmp(value, "none")) {
1582 vms->iommu = VIRT_IOMMU_NONE;
1583 } else {
1584 error_setg(errp, "Invalid iommu value");
1585 error_append_hint(errp, "Valid values are none, smmuv3.\n");
1586 }
1587 }
1588
1589 static CpuInstanceProperties
1590 virt_cpu_index_to_props(MachineState *ms, unsigned cpu_index)
1591 {
1592 MachineClass *mc = MACHINE_GET_CLASS(ms);
1593 const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms);
1594
1595 assert(cpu_index < possible_cpus->len);
1596 return possible_cpus->cpus[cpu_index].props;
1597 }
1598
1599 static int64_t virt_get_default_cpu_node_id(const MachineState *ms, int idx)
1600 {
1601 return idx % nb_numa_nodes;
1602 }
1603
1604 static const CPUArchIdList *virt_possible_cpu_arch_ids(MachineState *ms)
1605 {
1606 int n;
1607 VirtMachineState *vms = VIRT_MACHINE(ms);
1608
1609 if (ms->possible_cpus) {
1610 assert(ms->possible_cpus->len == max_cpus);
1611 return ms->possible_cpus;
1612 }
1613
1614 ms->possible_cpus = g_malloc0(sizeof(CPUArchIdList) +
1615 sizeof(CPUArchId) * max_cpus);
1616 ms->possible_cpus->len = max_cpus;
1617 for (n = 0; n < ms->possible_cpus->len; n++) {
1618 ms->possible_cpus->cpus[n].type = ms->cpu_type;
1619 ms->possible_cpus->cpus[n].arch_id =
1620 virt_cpu_mp_affinity(vms, n);
1621 ms->possible_cpus->cpus[n].props.has_thread_id = true;
1622 ms->possible_cpus->cpus[n].props.thread_id = n;
1623 }
1624 return ms->possible_cpus;
1625 }
1626
1627 static void virt_machine_device_plug_cb(HotplugHandler *hotplug_dev,
1628 DeviceState *dev, Error **errp)
1629 {
1630 VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
1631
1632 if (vms->platform_bus_dev) {
1633 if (object_dynamic_cast(OBJECT(dev), TYPE_SYS_BUS_DEVICE)) {
1634 platform_bus_link_device(PLATFORM_BUS_DEVICE(vms->platform_bus_dev),
1635 SYS_BUS_DEVICE(dev));
1636 }
1637 }
1638 }
1639
1640 static HotplugHandler *virt_machine_get_hotplug_handler(MachineState *machine,
1641 DeviceState *dev)
1642 {
1643 if (object_dynamic_cast(OBJECT(dev), TYPE_SYS_BUS_DEVICE)) {
1644 return HOTPLUG_HANDLER(machine);
1645 }
1646
1647 return NULL;
1648 }
1649
1650 static void virt_machine_class_init(ObjectClass *oc, void *data)
1651 {
1652 MachineClass *mc = MACHINE_CLASS(oc);
1653 HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
1654
1655 mc->init = machvirt_init;
1656 /* Start max_cpus at the maximum QEMU supports. We'll further restrict
1657 * it later in machvirt_init, where we have more information about the
1658 * configuration of the particular instance.
1659 */
1660 mc->max_cpus = 255;
1661 machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_CALXEDA_XGMAC);
1662 machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_AMD_XGBE);
1663 machine_class_allow_dynamic_sysbus_dev(mc, TYPE_RAMFB_DEVICE);
1664 mc->block_default_type = IF_VIRTIO;
1665 mc->no_cdrom = 1;
1666 mc->pci_allow_0_address = true;
1667 /* We know we will never create a pre-ARMv7 CPU which needs 1K pages */
1668 mc->minimum_page_bits = 12;
1669 mc->possible_cpu_arch_ids = virt_possible_cpu_arch_ids;
1670 mc->cpu_index_to_instance_props = virt_cpu_index_to_props;
1671 mc->default_cpu_type = ARM_CPU_TYPE_NAME("cortex-a15");
1672 mc->get_default_cpu_node_id = virt_get_default_cpu_node_id;
1673 assert(!mc->get_hotplug_handler);
1674 mc->get_hotplug_handler = virt_machine_get_hotplug_handler;
1675 hc->plug = virt_machine_device_plug_cb;
1676 }
1677
1678 static const TypeInfo virt_machine_info = {
1679 .name = TYPE_VIRT_MACHINE,
1680 .parent = TYPE_MACHINE,
1681 .abstract = true,
1682 .instance_size = sizeof(VirtMachineState),
1683 .class_size = sizeof(VirtMachineClass),
1684 .class_init = virt_machine_class_init,
1685 .interfaces = (InterfaceInfo[]) {
1686 { TYPE_HOTPLUG_HANDLER },
1687 { }
1688 },
1689 };
1690
1691 static void machvirt_machine_init(void)
1692 {
1693 type_register_static(&virt_machine_info);
1694 }
1695 type_init(machvirt_machine_init);
1696
1697 #define VIRT_COMPAT_2_12 \
1698 HW_COMPAT_2_12
1699
1700 static void virt_2_12_instance_init(Object *obj)
1701 {
1702 VirtMachineState *vms = VIRT_MACHINE(obj);
1703 VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1704
1705 /* EL3 is disabled by default on virt: this makes us consistent
1706 * between KVM and TCG for this board, and it also allows us to
1707 * boot UEFI blobs which assume no TrustZone support.
1708 */
1709 vms->secure = false;
1710 object_property_add_bool(obj, "secure", virt_get_secure,
1711 virt_set_secure, NULL);
1712 object_property_set_description(obj, "secure",
1713 "Set on/off to enable/disable the ARM "
1714 "Security Extensions (TrustZone)",
1715 NULL);
1716
1717 /* EL2 is also disabled by default, for similar reasons */
1718 vms->virt = false;
1719 object_property_add_bool(obj, "virtualization", virt_get_virt,
1720 virt_set_virt, NULL);
1721 object_property_set_description(obj, "virtualization",
1722 "Set on/off to enable/disable emulating a "
1723 "guest CPU which implements the ARM "
1724 "Virtualization Extensions",
1725 NULL);
1726
1727 /* High memory is enabled by default */
1728 vms->highmem = true;
1729 object_property_add_bool(obj, "highmem", virt_get_highmem,
1730 virt_set_highmem, NULL);
1731 object_property_set_description(obj, "highmem",
1732 "Set on/off to enable/disable using "
1733 "physical address space above 32 bits",
1734 NULL);
1735 /* Default GIC type is v2 */
1736 vms->gic_version = 2;
1737 object_property_add_str(obj, "gic-version", virt_get_gic_version,
1738 virt_set_gic_version, NULL);
1739 object_property_set_description(obj, "gic-version",
1740 "Set GIC version. "
1741 "Valid values are 2, 3 and host", NULL);
1742
1743 if (vmc->no_its) {
1744 vms->its = false;
1745 } else {
1746 /* Default allows ITS instantiation */
1747 vms->its = true;
1748 object_property_add_bool(obj, "its", virt_get_its,
1749 virt_set_its, NULL);
1750 object_property_set_description(obj, "its",
1751 "Set on/off to enable/disable "
1752 "ITS instantiation",
1753 NULL);
1754 }
1755
1756 /* Default disallows iommu instantiation */
1757 vms->iommu = VIRT_IOMMU_NONE;
1758 object_property_add_str(obj, "iommu", virt_get_iommu, virt_set_iommu, NULL);
1759 object_property_set_description(obj, "iommu",
1760 "Set the IOMMU type. "
1761 "Valid values are none and smmuv3",
1762 NULL);
1763
1764 vms->memmap = a15memmap;
1765 vms->irqmap = a15irqmap;
1766 }
1767
1768 static void virt_machine_2_12_options(MachineClass *mc)
1769 {
1770 SET_MACHINE_COMPAT(mc, VIRT_COMPAT_2_12);
1771 }
1772 DEFINE_VIRT_MACHINE_AS_LATEST(2, 12)
1773
1774 #define VIRT_COMPAT_2_11 \
1775 HW_COMPAT_2_11
1776
1777 static void virt_2_11_instance_init(Object *obj)
1778 {
1779 virt_2_12_instance_init(obj);
1780 }
1781
1782 static void virt_machine_2_11_options(MachineClass *mc)
1783 {
1784 VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
1785
1786 virt_machine_2_12_options(mc);
1787 SET_MACHINE_COMPAT(mc, VIRT_COMPAT_2_11);
1788 vmc->smbios_old_sys_ver = true;
1789 }
1790 DEFINE_VIRT_MACHINE(2, 11)
1791
1792 #define VIRT_COMPAT_2_10 \
1793 HW_COMPAT_2_10
1794
1795 static void virt_2_10_instance_init(Object *obj)
1796 {
1797 virt_2_11_instance_init(obj);
1798 }
1799
1800 static void virt_machine_2_10_options(MachineClass *mc)
1801 {
1802 virt_machine_2_11_options(mc);
1803 SET_MACHINE_COMPAT(mc, VIRT_COMPAT_2_10);
1804 }
1805 DEFINE_VIRT_MACHINE(2, 10)
1806
1807 #define VIRT_COMPAT_2_9 \
1808 HW_COMPAT_2_9
1809
1810 static void virt_2_9_instance_init(Object *obj)
1811 {
1812 virt_2_10_instance_init(obj);
1813 }
1814
1815 static void virt_machine_2_9_options(MachineClass *mc)
1816 {
1817 virt_machine_2_10_options(mc);
1818 SET_MACHINE_COMPAT(mc, VIRT_COMPAT_2_9);
1819 }
1820 DEFINE_VIRT_MACHINE(2, 9)
1821
1822 #define VIRT_COMPAT_2_8 \
1823 HW_COMPAT_2_8
1824
1825 static void virt_2_8_instance_init(Object *obj)
1826 {
1827 virt_2_9_instance_init(obj);
1828 }
1829
1830 static void virt_machine_2_8_options(MachineClass *mc)
1831 {
1832 VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
1833
1834 virt_machine_2_9_options(mc);
1835 SET_MACHINE_COMPAT(mc, VIRT_COMPAT_2_8);
1836 /* For 2.8 and earlier we falsely claimed in the DT that
1837 * our timers were edge-triggered, not level-triggered.
1838 */
1839 vmc->claim_edge_triggered_timers = true;
1840 }
1841 DEFINE_VIRT_MACHINE(2, 8)
1842
1843 #define VIRT_COMPAT_2_7 \
1844 HW_COMPAT_2_7
1845
1846 static void virt_2_7_instance_init(Object *obj)
1847 {
1848 virt_2_8_instance_init(obj);
1849 }
1850
1851 static void virt_machine_2_7_options(MachineClass *mc)
1852 {
1853 VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
1854
1855 virt_machine_2_8_options(mc);
1856 SET_MACHINE_COMPAT(mc, VIRT_COMPAT_2_7);
1857 /* ITS was introduced with 2.8 */
1858 vmc->no_its = true;
1859 /* Stick with 1K pages for migration compatibility */
1860 mc->minimum_page_bits = 0;
1861 }
1862 DEFINE_VIRT_MACHINE(2, 7)
1863
1864 #define VIRT_COMPAT_2_6 \
1865 HW_COMPAT_2_6
1866
1867 static void virt_2_6_instance_init(Object *obj)
1868 {
1869 virt_2_7_instance_init(obj);
1870 }
1871
1872 static void virt_machine_2_6_options(MachineClass *mc)
1873 {
1874 VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
1875
1876 virt_machine_2_7_options(mc);
1877 SET_MACHINE_COMPAT(mc, VIRT_COMPAT_2_6);
1878 vmc->disallow_affinity_adjustment = true;
1879 /* Disable PMU for 2.6 as PMU support was first introduced in 2.7 */
1880 vmc->no_pmu = true;
1881 }
1882 DEFINE_VIRT_MACHINE(2, 6)