]> git.proxmox.com Git - mirror_qemu.git/blob - hw/arm/virt.c
Merge remote-tracking branch 'remotes/afaerber/tags/qom-devices-for-peter' 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 "hw/sysbus.h"
33 #include "hw/arm/arm.h"
34 #include "hw/arm/primecell.h"
35 #include "hw/arm/virt.h"
36 #include "hw/devices.h"
37 #include "net/net.h"
38 #include "sysemu/block-backend.h"
39 #include "sysemu/device_tree.h"
40 #include "sysemu/sysemu.h"
41 #include "sysemu/kvm.h"
42 #include "hw/boards.h"
43 #include "hw/loader.h"
44 #include "exec/address-spaces.h"
45 #include "qemu/bitops.h"
46 #include "qemu/error-report.h"
47 #include "hw/pci-host/gpex.h"
48 #include "hw/arm/virt-acpi-build.h"
49 #include "hw/arm/sysbus-fdt.h"
50 #include "hw/platform-bus.h"
51 #include "hw/arm/fdt.h"
52 #include "hw/intc/arm_gic_common.h"
53 #include "kvm_arm.h"
54 #include "hw/smbios/smbios.h"
55 #include "qapi/visitor.h"
56 #include "standard-headers/linux/input.h"
57
58 /* Number of external interrupt lines to configure the GIC with */
59 #define NUM_IRQS 256
60
61 #define PLATFORM_BUS_NUM_IRQS 64
62
63 static ARMPlatformBusSystemParams platform_bus_params;
64
65 typedef struct VirtBoardInfo {
66 struct arm_boot_info bootinfo;
67 const char *cpu_model;
68 const MemMapEntry *memmap;
69 const int *irqmap;
70 int smp_cpus;
71 void *fdt;
72 int fdt_size;
73 uint32_t clock_phandle;
74 uint32_t gic_phandle;
75 uint32_t v2m_phandle;
76 } VirtBoardInfo;
77
78 typedef struct {
79 MachineClass parent;
80 VirtBoardInfo *daughterboard;
81 } VirtMachineClass;
82
83 typedef struct {
84 MachineState parent;
85 bool secure;
86 bool highmem;
87 int32_t gic_version;
88 } VirtMachineState;
89
90 #define TYPE_VIRT_MACHINE MACHINE_TYPE_NAME("virt")
91 #define VIRT_MACHINE(obj) \
92 OBJECT_CHECK(VirtMachineState, (obj), TYPE_VIRT_MACHINE)
93 #define VIRT_MACHINE_GET_CLASS(obj) \
94 OBJECT_GET_CLASS(VirtMachineClass, obj, TYPE_VIRT_MACHINE)
95 #define VIRT_MACHINE_CLASS(klass) \
96 OBJECT_CLASS_CHECK(VirtMachineClass, klass, TYPE_VIRT_MACHINE)
97
98 /* Addresses and sizes of our components.
99 * 0..128MB is space for a flash device so we can run bootrom code such as UEFI.
100 * 128MB..256MB is used for miscellaneous device I/O.
101 * 256MB..1GB is reserved for possible future PCI support (ie where the
102 * PCI memory window will go if we add a PCI host controller).
103 * 1GB and up is RAM (which may happily spill over into the
104 * high memory region beyond 4GB).
105 * This represents a compromise between how much RAM can be given to
106 * a 32 bit VM and leaving space for expansion and in particular for PCI.
107 * Note that devices should generally be placed at multiples of 0x10000,
108 * to accommodate guests using 64K pages.
109 */
110 static const MemMapEntry a15memmap[] = {
111 /* Space up to 0x8000000 is reserved for a boot ROM */
112 [VIRT_FLASH] = { 0, 0x08000000 },
113 [VIRT_CPUPERIPHS] = { 0x08000000, 0x00020000 },
114 /* GIC distributor and CPU interfaces sit inside the CPU peripheral space */
115 [VIRT_GIC_DIST] = { 0x08000000, 0x00010000 },
116 [VIRT_GIC_CPU] = { 0x08010000, 0x00010000 },
117 [VIRT_GIC_V2M] = { 0x08020000, 0x00001000 },
118 /* The space in between here is reserved for GICv3 CPU/vCPU/HYP */
119 [VIRT_GIC_ITS] = { 0x08080000, 0x00020000 },
120 /* This redistributor space allows up to 2*64kB*123 CPUs */
121 [VIRT_GIC_REDIST] = { 0x080A0000, 0x00F60000 },
122 [VIRT_UART] = { 0x09000000, 0x00001000 },
123 [VIRT_RTC] = { 0x09010000, 0x00001000 },
124 [VIRT_FW_CFG] = { 0x09020000, 0x00000018 },
125 [VIRT_GPIO] = { 0x09030000, 0x00001000 },
126 [VIRT_MMIO] = { 0x0a000000, 0x00000200 },
127 /* ...repeating for a total of NUM_VIRTIO_TRANSPORTS, each of that size */
128 [VIRT_PLATFORM_BUS] = { 0x0c000000, 0x02000000 },
129 [VIRT_PCIE_MMIO] = { 0x10000000, 0x2eff0000 },
130 [VIRT_PCIE_PIO] = { 0x3eff0000, 0x00010000 },
131 [VIRT_PCIE_ECAM] = { 0x3f000000, 0x01000000 },
132 [VIRT_MEM] = { 0x40000000, 30ULL * 1024 * 1024 * 1024 },
133 /* Second PCIe window, 512GB wide at the 512GB boundary */
134 [VIRT_PCIE_MMIO_HIGH] = { 0x8000000000ULL, 0x8000000000ULL },
135 };
136
137 static const int a15irqmap[] = {
138 [VIRT_UART] = 1,
139 [VIRT_RTC] = 2,
140 [VIRT_PCIE] = 3, /* ... to 6 */
141 [VIRT_GPIO] = 7,
142 [VIRT_MMIO] = 16, /* ...to 16 + NUM_VIRTIO_TRANSPORTS - 1 */
143 [VIRT_GIC_V2M] = 48, /* ...to 48 + NUM_GICV2M_SPIS - 1 */
144 [VIRT_PLATFORM_BUS] = 112, /* ...to 112 + PLATFORM_BUS_NUM_IRQS -1 */
145 };
146
147 static VirtBoardInfo machines[] = {
148 {
149 .cpu_model = "cortex-a15",
150 .memmap = a15memmap,
151 .irqmap = a15irqmap,
152 },
153 {
154 .cpu_model = "cortex-a53",
155 .memmap = a15memmap,
156 .irqmap = a15irqmap,
157 },
158 {
159 .cpu_model = "cortex-a57",
160 .memmap = a15memmap,
161 .irqmap = a15irqmap,
162 },
163 {
164 .cpu_model = "host",
165 .memmap = a15memmap,
166 .irqmap = a15irqmap,
167 },
168 };
169
170 static VirtBoardInfo *find_machine_info(const char *cpu)
171 {
172 int i;
173
174 for (i = 0; i < ARRAY_SIZE(machines); i++) {
175 if (strcmp(cpu, machines[i].cpu_model) == 0) {
176 return &machines[i];
177 }
178 }
179 return NULL;
180 }
181
182 static void create_fdt(VirtBoardInfo *vbi)
183 {
184 void *fdt = create_device_tree(&vbi->fdt_size);
185
186 if (!fdt) {
187 error_report("create_device_tree() failed");
188 exit(1);
189 }
190
191 vbi->fdt = fdt;
192
193 /* Header */
194 qemu_fdt_setprop_string(fdt, "/", "compatible", "linux,dummy-virt");
195 qemu_fdt_setprop_cell(fdt, "/", "#address-cells", 0x2);
196 qemu_fdt_setprop_cell(fdt, "/", "#size-cells", 0x2);
197
198 /*
199 * /chosen and /memory nodes must exist for load_dtb
200 * to fill in necessary properties later
201 */
202 qemu_fdt_add_subnode(fdt, "/chosen");
203 qemu_fdt_add_subnode(fdt, "/memory");
204 qemu_fdt_setprop_string(fdt, "/memory", "device_type", "memory");
205
206 /* Clock node, for the benefit of the UART. The kernel device tree
207 * binding documentation claims the PL011 node clock properties are
208 * optional but in practice if you omit them the kernel refuses to
209 * probe for the device.
210 */
211 vbi->clock_phandle = qemu_fdt_alloc_phandle(fdt);
212 qemu_fdt_add_subnode(fdt, "/apb-pclk");
213 qemu_fdt_setprop_string(fdt, "/apb-pclk", "compatible", "fixed-clock");
214 qemu_fdt_setprop_cell(fdt, "/apb-pclk", "#clock-cells", 0x0);
215 qemu_fdt_setprop_cell(fdt, "/apb-pclk", "clock-frequency", 24000000);
216 qemu_fdt_setprop_string(fdt, "/apb-pclk", "clock-output-names",
217 "clk24mhz");
218 qemu_fdt_setprop_cell(fdt, "/apb-pclk", "phandle", vbi->clock_phandle);
219
220 }
221
222 static void fdt_add_psci_node(const VirtBoardInfo *vbi)
223 {
224 uint32_t cpu_suspend_fn;
225 uint32_t cpu_off_fn;
226 uint32_t cpu_on_fn;
227 uint32_t migrate_fn;
228 void *fdt = vbi->fdt;
229 ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(0));
230
231 qemu_fdt_add_subnode(fdt, "/psci");
232 if (armcpu->psci_version == 2) {
233 const char comp[] = "arm,psci-0.2\0arm,psci";
234 qemu_fdt_setprop(fdt, "/psci", "compatible", comp, sizeof(comp));
235
236 cpu_off_fn = QEMU_PSCI_0_2_FN_CPU_OFF;
237 if (arm_feature(&armcpu->env, ARM_FEATURE_AARCH64)) {
238 cpu_suspend_fn = QEMU_PSCI_0_2_FN64_CPU_SUSPEND;
239 cpu_on_fn = QEMU_PSCI_0_2_FN64_CPU_ON;
240 migrate_fn = QEMU_PSCI_0_2_FN64_MIGRATE;
241 } else {
242 cpu_suspend_fn = QEMU_PSCI_0_2_FN_CPU_SUSPEND;
243 cpu_on_fn = QEMU_PSCI_0_2_FN_CPU_ON;
244 migrate_fn = QEMU_PSCI_0_2_FN_MIGRATE;
245 }
246 } else {
247 qemu_fdt_setprop_string(fdt, "/psci", "compatible", "arm,psci");
248
249 cpu_suspend_fn = QEMU_PSCI_0_1_FN_CPU_SUSPEND;
250 cpu_off_fn = QEMU_PSCI_0_1_FN_CPU_OFF;
251 cpu_on_fn = QEMU_PSCI_0_1_FN_CPU_ON;
252 migrate_fn = QEMU_PSCI_0_1_FN_MIGRATE;
253 }
254
255 /* We adopt the PSCI spec's nomenclature, and use 'conduit' to refer
256 * to the instruction that should be used to invoke PSCI functions.
257 * However, the device tree binding uses 'method' instead, so that is
258 * what we should use here.
259 */
260 qemu_fdt_setprop_string(fdt, "/psci", "method", "hvc");
261
262 qemu_fdt_setprop_cell(fdt, "/psci", "cpu_suspend", cpu_suspend_fn);
263 qemu_fdt_setprop_cell(fdt, "/psci", "cpu_off", cpu_off_fn);
264 qemu_fdt_setprop_cell(fdt, "/psci", "cpu_on", cpu_on_fn);
265 qemu_fdt_setprop_cell(fdt, "/psci", "migrate", migrate_fn);
266 }
267
268 static void fdt_add_timer_nodes(const VirtBoardInfo *vbi, int gictype)
269 {
270 /* Note that on A15 h/w these interrupts are level-triggered,
271 * but for the GIC implementation provided by both QEMU and KVM
272 * they are edge-triggered.
273 */
274 ARMCPU *armcpu;
275 uint32_t irqflags = GIC_FDT_IRQ_FLAGS_EDGE_LO_HI;
276
277 if (gictype == 2) {
278 irqflags = deposit32(irqflags, GIC_FDT_IRQ_PPI_CPU_START,
279 GIC_FDT_IRQ_PPI_CPU_WIDTH,
280 (1 << vbi->smp_cpus) - 1);
281 }
282
283 qemu_fdt_add_subnode(vbi->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(vbi->fdt, "/timer", "compatible",
289 compat, sizeof(compat));
290 } else {
291 qemu_fdt_setprop_string(vbi->fdt, "/timer", "compatible",
292 "arm,armv7-timer");
293 }
294 qemu_fdt_setprop_cells(vbi->fdt, "/timer", "interrupts",
295 GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_S_EL1_IRQ, irqflags,
296 GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_NS_EL1_IRQ, irqflags,
297 GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_VIRT_IRQ, irqflags,
298 GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_NS_EL2_IRQ, irqflags);
299 }
300
301 static void fdt_add_cpu_nodes(const VirtBoardInfo *vbi)
302 {
303 int cpu;
304 int addr_cells = 1;
305
306 /*
307 * From Documentation/devicetree/bindings/arm/cpus.txt
308 * On ARM v8 64-bit systems value should be set to 2,
309 * that corresponds to the MPIDR_EL1 register size.
310 * If MPIDR_EL1[63:32] value is equal to 0 on all CPUs
311 * in the system, #address-cells can be set to 1, since
312 * MPIDR_EL1[63:32] bits are not used for CPUs
313 * identification.
314 *
315 * Here we actually don't know whether our system is 32- or 64-bit one.
316 * The simplest way to go is to examine affinity IDs of all our CPUs. If
317 * at least one of them has Aff3 populated, we set #address-cells to 2.
318 */
319 for (cpu = 0; cpu < vbi->smp_cpus; cpu++) {
320 ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
321
322 if (armcpu->mp_affinity & ARM_AFF3_MASK) {
323 addr_cells = 2;
324 break;
325 }
326 }
327
328 qemu_fdt_add_subnode(vbi->fdt, "/cpus");
329 qemu_fdt_setprop_cell(vbi->fdt, "/cpus", "#address-cells", addr_cells);
330 qemu_fdt_setprop_cell(vbi->fdt, "/cpus", "#size-cells", 0x0);
331
332 for (cpu = vbi->smp_cpus - 1; cpu >= 0; cpu--) {
333 char *nodename = g_strdup_printf("/cpus/cpu@%d", cpu);
334 ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
335
336 qemu_fdt_add_subnode(vbi->fdt, nodename);
337 qemu_fdt_setprop_string(vbi->fdt, nodename, "device_type", "cpu");
338 qemu_fdt_setprop_string(vbi->fdt, nodename, "compatible",
339 armcpu->dtb_compatible);
340
341 if (vbi->smp_cpus > 1) {
342 qemu_fdt_setprop_string(vbi->fdt, nodename,
343 "enable-method", "psci");
344 }
345
346 if (addr_cells == 2) {
347 qemu_fdt_setprop_u64(vbi->fdt, nodename, "reg",
348 armcpu->mp_affinity);
349 } else {
350 qemu_fdt_setprop_cell(vbi->fdt, nodename, "reg",
351 armcpu->mp_affinity);
352 }
353
354 g_free(nodename);
355 }
356 }
357
358 static void fdt_add_v2m_gic_node(VirtBoardInfo *vbi)
359 {
360 vbi->v2m_phandle = qemu_fdt_alloc_phandle(vbi->fdt);
361 qemu_fdt_add_subnode(vbi->fdt, "/intc/v2m");
362 qemu_fdt_setprop_string(vbi->fdt, "/intc/v2m", "compatible",
363 "arm,gic-v2m-frame");
364 qemu_fdt_setprop(vbi->fdt, "/intc/v2m", "msi-controller", NULL, 0);
365 qemu_fdt_setprop_sized_cells(vbi->fdt, "/intc/v2m", "reg",
366 2, vbi->memmap[VIRT_GIC_V2M].base,
367 2, vbi->memmap[VIRT_GIC_V2M].size);
368 qemu_fdt_setprop_cell(vbi->fdt, "/intc/v2m", "phandle", vbi->v2m_phandle);
369 }
370
371 static void fdt_add_gic_node(VirtBoardInfo *vbi, int type)
372 {
373 vbi->gic_phandle = qemu_fdt_alloc_phandle(vbi->fdt);
374 qemu_fdt_setprop_cell(vbi->fdt, "/", "interrupt-parent", vbi->gic_phandle);
375
376 qemu_fdt_add_subnode(vbi->fdt, "/intc");
377 qemu_fdt_setprop_cell(vbi->fdt, "/intc", "#interrupt-cells", 3);
378 qemu_fdt_setprop(vbi->fdt, "/intc", "interrupt-controller", NULL, 0);
379 qemu_fdt_setprop_cell(vbi->fdt, "/intc", "#address-cells", 0x2);
380 qemu_fdt_setprop_cell(vbi->fdt, "/intc", "#size-cells", 0x2);
381 qemu_fdt_setprop(vbi->fdt, "/intc", "ranges", NULL, 0);
382 if (type == 3) {
383 qemu_fdt_setprop_string(vbi->fdt, "/intc", "compatible",
384 "arm,gic-v3");
385 qemu_fdt_setprop_sized_cells(vbi->fdt, "/intc", "reg",
386 2, vbi->memmap[VIRT_GIC_DIST].base,
387 2, vbi->memmap[VIRT_GIC_DIST].size,
388 2, vbi->memmap[VIRT_GIC_REDIST].base,
389 2, vbi->memmap[VIRT_GIC_REDIST].size);
390 } else {
391 /* 'cortex-a15-gic' means 'GIC v2' */
392 qemu_fdt_setprop_string(vbi->fdt, "/intc", "compatible",
393 "arm,cortex-a15-gic");
394 qemu_fdt_setprop_sized_cells(vbi->fdt, "/intc", "reg",
395 2, vbi->memmap[VIRT_GIC_DIST].base,
396 2, vbi->memmap[VIRT_GIC_DIST].size,
397 2, vbi->memmap[VIRT_GIC_CPU].base,
398 2, vbi->memmap[VIRT_GIC_CPU].size);
399 }
400
401 qemu_fdt_setprop_cell(vbi->fdt, "/intc", "phandle", vbi->gic_phandle);
402 }
403
404 static void create_v2m(VirtBoardInfo *vbi, qemu_irq *pic)
405 {
406 int i;
407 int irq = vbi->irqmap[VIRT_GIC_V2M];
408 DeviceState *dev;
409
410 dev = qdev_create(NULL, "arm-gicv2m");
411 sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vbi->memmap[VIRT_GIC_V2M].base);
412 qdev_prop_set_uint32(dev, "base-spi", irq);
413 qdev_prop_set_uint32(dev, "num-spi", NUM_GICV2M_SPIS);
414 qdev_init_nofail(dev);
415
416 for (i = 0; i < NUM_GICV2M_SPIS; i++) {
417 sysbus_connect_irq(SYS_BUS_DEVICE(dev), i, pic[irq + i]);
418 }
419
420 fdt_add_v2m_gic_node(vbi);
421 }
422
423 static void create_gic(VirtBoardInfo *vbi, qemu_irq *pic, int type, bool secure)
424 {
425 /* We create a standalone GIC */
426 DeviceState *gicdev;
427 SysBusDevice *gicbusdev;
428 const char *gictype;
429 int i;
430
431 gictype = (type == 3) ? gicv3_class_name() : gic_class_name();
432
433 gicdev = qdev_create(NULL, gictype);
434 qdev_prop_set_uint32(gicdev, "revision", type);
435 qdev_prop_set_uint32(gicdev, "num-cpu", smp_cpus);
436 /* Note that the num-irq property counts both internal and external
437 * interrupts; there are always 32 of the former (mandated by GIC spec).
438 */
439 qdev_prop_set_uint32(gicdev, "num-irq", NUM_IRQS + 32);
440 if (!kvm_irqchip_in_kernel()) {
441 qdev_prop_set_bit(gicdev, "has-security-extensions", secure);
442 }
443 qdev_init_nofail(gicdev);
444 gicbusdev = SYS_BUS_DEVICE(gicdev);
445 sysbus_mmio_map(gicbusdev, 0, vbi->memmap[VIRT_GIC_DIST].base);
446 if (type == 3) {
447 sysbus_mmio_map(gicbusdev, 1, vbi->memmap[VIRT_GIC_REDIST].base);
448 } else {
449 sysbus_mmio_map(gicbusdev, 1, vbi->memmap[VIRT_GIC_CPU].base);
450 }
451
452 /* Wire the outputs from each CPU's generic timer to the
453 * appropriate GIC PPI inputs, and the GIC's IRQ output to
454 * the CPU's IRQ input.
455 */
456 for (i = 0; i < smp_cpus; i++) {
457 DeviceState *cpudev = DEVICE(qemu_get_cpu(i));
458 int ppibase = NUM_IRQS + i * GIC_INTERNAL + GIC_NR_SGIS;
459 int irq;
460 /* Mapping from the output timer irq lines from the CPU to the
461 * GIC PPI inputs we use for the virt board.
462 */
463 const int timer_irq[] = {
464 [GTIMER_PHYS] = ARCH_TIMER_NS_EL1_IRQ,
465 [GTIMER_VIRT] = ARCH_TIMER_VIRT_IRQ,
466 [GTIMER_HYP] = ARCH_TIMER_NS_EL2_IRQ,
467 [GTIMER_SEC] = ARCH_TIMER_S_EL1_IRQ,
468 };
469
470 for (irq = 0; irq < ARRAY_SIZE(timer_irq); irq++) {
471 qdev_connect_gpio_out(cpudev, irq,
472 qdev_get_gpio_in(gicdev,
473 ppibase + timer_irq[irq]));
474 }
475
476 sysbus_connect_irq(gicbusdev, i, qdev_get_gpio_in(cpudev, ARM_CPU_IRQ));
477 sysbus_connect_irq(gicbusdev, i + smp_cpus,
478 qdev_get_gpio_in(cpudev, ARM_CPU_FIQ));
479 }
480
481 for (i = 0; i < NUM_IRQS; i++) {
482 pic[i] = qdev_get_gpio_in(gicdev, i);
483 }
484
485 fdt_add_gic_node(vbi, type);
486
487 if (type == 2) {
488 create_v2m(vbi, pic);
489 }
490 }
491
492 static void create_uart(const VirtBoardInfo *vbi, qemu_irq *pic)
493 {
494 char *nodename;
495 hwaddr base = vbi->memmap[VIRT_UART].base;
496 hwaddr size = vbi->memmap[VIRT_UART].size;
497 int irq = vbi->irqmap[VIRT_UART];
498 const char compat[] = "arm,pl011\0arm,primecell";
499 const char clocknames[] = "uartclk\0apb_pclk";
500
501 sysbus_create_simple("pl011", base, pic[irq]);
502
503 nodename = g_strdup_printf("/pl011@%" PRIx64, base);
504 qemu_fdt_add_subnode(vbi->fdt, nodename);
505 /* Note that we can't use setprop_string because of the embedded NUL */
506 qemu_fdt_setprop(vbi->fdt, nodename, "compatible",
507 compat, sizeof(compat));
508 qemu_fdt_setprop_sized_cells(vbi->fdt, nodename, "reg",
509 2, base, 2, size);
510 qemu_fdt_setprop_cells(vbi->fdt, nodename, "interrupts",
511 GIC_FDT_IRQ_TYPE_SPI, irq,
512 GIC_FDT_IRQ_FLAGS_LEVEL_HI);
513 qemu_fdt_setprop_cells(vbi->fdt, nodename, "clocks",
514 vbi->clock_phandle, vbi->clock_phandle);
515 qemu_fdt_setprop(vbi->fdt, nodename, "clock-names",
516 clocknames, sizeof(clocknames));
517
518 qemu_fdt_setprop_string(vbi->fdt, "/chosen", "stdout-path", nodename);
519 g_free(nodename);
520 }
521
522 static void create_rtc(const VirtBoardInfo *vbi, qemu_irq *pic)
523 {
524 char *nodename;
525 hwaddr base = vbi->memmap[VIRT_RTC].base;
526 hwaddr size = vbi->memmap[VIRT_RTC].size;
527 int irq = vbi->irqmap[VIRT_RTC];
528 const char compat[] = "arm,pl031\0arm,primecell";
529
530 sysbus_create_simple("pl031", base, pic[irq]);
531
532 nodename = g_strdup_printf("/pl031@%" PRIx64, base);
533 qemu_fdt_add_subnode(vbi->fdt, nodename);
534 qemu_fdt_setprop(vbi->fdt, nodename, "compatible", compat, sizeof(compat));
535 qemu_fdt_setprop_sized_cells(vbi->fdt, nodename, "reg",
536 2, base, 2, size);
537 qemu_fdt_setprop_cells(vbi->fdt, nodename, "interrupts",
538 GIC_FDT_IRQ_TYPE_SPI, irq,
539 GIC_FDT_IRQ_FLAGS_LEVEL_HI);
540 qemu_fdt_setprop_cell(vbi->fdt, nodename, "clocks", vbi->clock_phandle);
541 qemu_fdt_setprop_string(vbi->fdt, nodename, "clock-names", "apb_pclk");
542 g_free(nodename);
543 }
544
545 static DeviceState *pl061_dev;
546 static void virt_powerdown_req(Notifier *n, void *opaque)
547 {
548 /* use gpio Pin 3 for power button event */
549 qemu_set_irq(qdev_get_gpio_in(pl061_dev, 3), 1);
550 }
551
552 static Notifier virt_system_powerdown_notifier = {
553 .notify = virt_powerdown_req
554 };
555
556 static void create_gpio(const VirtBoardInfo *vbi, qemu_irq *pic)
557 {
558 char *nodename;
559 hwaddr base = vbi->memmap[VIRT_GPIO].base;
560 hwaddr size = vbi->memmap[VIRT_GPIO].size;
561 int irq = vbi->irqmap[VIRT_GPIO];
562 const char compat[] = "arm,pl061\0arm,primecell";
563
564 pl061_dev = sysbus_create_simple("pl061", base, pic[irq]);
565
566 uint32_t phandle = qemu_fdt_alloc_phandle(vbi->fdt);
567 nodename = g_strdup_printf("/pl061@%" PRIx64, base);
568 qemu_fdt_add_subnode(vbi->fdt, nodename);
569 qemu_fdt_setprop_sized_cells(vbi->fdt, nodename, "reg",
570 2, base, 2, size);
571 qemu_fdt_setprop(vbi->fdt, nodename, "compatible", compat, sizeof(compat));
572 qemu_fdt_setprop_cell(vbi->fdt, nodename, "#gpio-cells", 2);
573 qemu_fdt_setprop(vbi->fdt, nodename, "gpio-controller", NULL, 0);
574 qemu_fdt_setprop_cells(vbi->fdt, nodename, "interrupts",
575 GIC_FDT_IRQ_TYPE_SPI, irq,
576 GIC_FDT_IRQ_FLAGS_LEVEL_HI);
577 qemu_fdt_setprop_cell(vbi->fdt, nodename, "clocks", vbi->clock_phandle);
578 qemu_fdt_setprop_string(vbi->fdt, nodename, "clock-names", "apb_pclk");
579 qemu_fdt_setprop_cell(vbi->fdt, nodename, "phandle", phandle);
580
581 qemu_fdt_add_subnode(vbi->fdt, "/gpio-keys");
582 qemu_fdt_setprop_string(vbi->fdt, "/gpio-keys", "compatible", "gpio-keys");
583 qemu_fdt_setprop_cell(vbi->fdt, "/gpio-keys", "#size-cells", 0);
584 qemu_fdt_setprop_cell(vbi->fdt, "/gpio-keys", "#address-cells", 1);
585
586 qemu_fdt_add_subnode(vbi->fdt, "/gpio-keys/poweroff");
587 qemu_fdt_setprop_string(vbi->fdt, "/gpio-keys/poweroff",
588 "label", "GPIO Key Poweroff");
589 qemu_fdt_setprop_cell(vbi->fdt, "/gpio-keys/poweroff", "linux,code",
590 KEY_POWER);
591 qemu_fdt_setprop_cells(vbi->fdt, "/gpio-keys/poweroff",
592 "gpios", phandle, 3, 0);
593
594 /* connect powerdown request */
595 qemu_register_powerdown_notifier(&virt_system_powerdown_notifier);
596
597 g_free(nodename);
598 }
599
600 static void create_virtio_devices(const VirtBoardInfo *vbi, qemu_irq *pic)
601 {
602 int i;
603 hwaddr size = vbi->memmap[VIRT_MMIO].size;
604
605 /* We create the transports in forwards order. Since qbus_realize()
606 * prepends (not appends) new child buses, the incrementing loop below will
607 * create a list of virtio-mmio buses with decreasing base addresses.
608 *
609 * When a -device option is processed from the command line,
610 * qbus_find_recursive() picks the next free virtio-mmio bus in forwards
611 * order. The upshot is that -device options in increasing command line
612 * order are mapped to virtio-mmio buses with decreasing base addresses.
613 *
614 * When this code was originally written, that arrangement ensured that the
615 * guest Linux kernel would give the lowest "name" (/dev/vda, eth0, etc) to
616 * the first -device on the command line. (The end-to-end order is a
617 * function of this loop, qbus_realize(), qbus_find_recursive(), and the
618 * guest kernel's name-to-address assignment strategy.)
619 *
620 * Meanwhile, the kernel's traversal seems to have been reversed; see eg.
621 * the message, if not necessarily the code, of commit 70161ff336.
622 * Therefore the loop now establishes the inverse of the original intent.
623 *
624 * Unfortunately, we can't counteract the kernel change by reversing the
625 * loop; it would break existing command lines.
626 *
627 * In any case, the kernel makes no guarantee about the stability of
628 * enumeration order of virtio devices (as demonstrated by it changing
629 * between kernel versions). For reliable and stable identification
630 * of disks users must use UUIDs or similar mechanisms.
631 */
632 for (i = 0; i < NUM_VIRTIO_TRANSPORTS; i++) {
633 int irq = vbi->irqmap[VIRT_MMIO] + i;
634 hwaddr base = vbi->memmap[VIRT_MMIO].base + i * size;
635
636 sysbus_create_simple("virtio-mmio", base, pic[irq]);
637 }
638
639 /* We add dtb nodes in reverse order so that they appear in the finished
640 * device tree lowest address first.
641 *
642 * Note that this mapping is independent of the loop above. The previous
643 * loop influences virtio device to virtio transport assignment, whereas
644 * this loop controls how virtio transports are laid out in the dtb.
645 */
646 for (i = NUM_VIRTIO_TRANSPORTS - 1; i >= 0; i--) {
647 char *nodename;
648 int irq = vbi->irqmap[VIRT_MMIO] + i;
649 hwaddr base = vbi->memmap[VIRT_MMIO].base + i * size;
650
651 nodename = g_strdup_printf("/virtio_mmio@%" PRIx64, base);
652 qemu_fdt_add_subnode(vbi->fdt, nodename);
653 qemu_fdt_setprop_string(vbi->fdt, nodename,
654 "compatible", "virtio,mmio");
655 qemu_fdt_setprop_sized_cells(vbi->fdt, nodename, "reg",
656 2, base, 2, size);
657 qemu_fdt_setprop_cells(vbi->fdt, nodename, "interrupts",
658 GIC_FDT_IRQ_TYPE_SPI, irq,
659 GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
660 g_free(nodename);
661 }
662 }
663
664 static void create_one_flash(const char *name, hwaddr flashbase,
665 hwaddr flashsize)
666 {
667 /* Create and map a single flash device. We use the same
668 * parameters as the flash devices on the Versatile Express board.
669 */
670 DriveInfo *dinfo = drive_get_next(IF_PFLASH);
671 DeviceState *dev = qdev_create(NULL, "cfi.pflash01");
672 const uint64_t sectorlength = 256 * 1024;
673
674 if (dinfo) {
675 qdev_prop_set_drive(dev, "drive", blk_by_legacy_dinfo(dinfo),
676 &error_abort);
677 }
678
679 qdev_prop_set_uint32(dev, "num-blocks", flashsize / sectorlength);
680 qdev_prop_set_uint64(dev, "sector-length", sectorlength);
681 qdev_prop_set_uint8(dev, "width", 4);
682 qdev_prop_set_uint8(dev, "device-width", 2);
683 qdev_prop_set_bit(dev, "big-endian", false);
684 qdev_prop_set_uint16(dev, "id0", 0x89);
685 qdev_prop_set_uint16(dev, "id1", 0x18);
686 qdev_prop_set_uint16(dev, "id2", 0x00);
687 qdev_prop_set_uint16(dev, "id3", 0x00);
688 qdev_prop_set_string(dev, "name", name);
689 qdev_init_nofail(dev);
690
691 sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, flashbase);
692 }
693
694 static void create_flash(const VirtBoardInfo *vbi)
695 {
696 /* Create two flash devices to fill the VIRT_FLASH space in the memmap.
697 * Any file passed via -bios goes in the first of these.
698 */
699 hwaddr flashsize = vbi->memmap[VIRT_FLASH].size / 2;
700 hwaddr flashbase = vbi->memmap[VIRT_FLASH].base;
701 char *nodename;
702
703 if (bios_name) {
704 char *fn;
705 int image_size;
706
707 if (drive_get(IF_PFLASH, 0, 0)) {
708 error_report("The contents of the first flash device may be "
709 "specified with -bios or with -drive if=pflash... "
710 "but you cannot use both options at once");
711 exit(1);
712 }
713 fn = qemu_find_file(QEMU_FILE_TYPE_BIOS, bios_name);
714 if (!fn) {
715 error_report("Could not find ROM image '%s'", bios_name);
716 exit(1);
717 }
718 image_size = load_image_targphys(fn, flashbase, flashsize);
719 g_free(fn);
720 if (image_size < 0) {
721 error_report("Could not load ROM image '%s'", bios_name);
722 exit(1);
723 }
724 }
725
726 create_one_flash("virt.flash0", flashbase, flashsize);
727 create_one_flash("virt.flash1", flashbase + flashsize, flashsize);
728
729 nodename = g_strdup_printf("/flash@%" PRIx64, flashbase);
730 qemu_fdt_add_subnode(vbi->fdt, nodename);
731 qemu_fdt_setprop_string(vbi->fdt, nodename, "compatible", "cfi-flash");
732 qemu_fdt_setprop_sized_cells(vbi->fdt, nodename, "reg",
733 2, flashbase, 2, flashsize,
734 2, flashbase + flashsize, 2, flashsize);
735 qemu_fdt_setprop_cell(vbi->fdt, nodename, "bank-width", 4);
736 g_free(nodename);
737 }
738
739 static void create_fw_cfg(const VirtBoardInfo *vbi, AddressSpace *as)
740 {
741 hwaddr base = vbi->memmap[VIRT_FW_CFG].base;
742 hwaddr size = vbi->memmap[VIRT_FW_CFG].size;
743 char *nodename;
744
745 fw_cfg_init_mem_wide(base + 8, base, 8, base + 16, as);
746
747 nodename = g_strdup_printf("/fw-cfg@%" PRIx64, base);
748 qemu_fdt_add_subnode(vbi->fdt, nodename);
749 qemu_fdt_setprop_string(vbi->fdt, nodename,
750 "compatible", "qemu,fw-cfg-mmio");
751 qemu_fdt_setprop_sized_cells(vbi->fdt, nodename, "reg",
752 2, base, 2, size);
753 g_free(nodename);
754 }
755
756 static void create_pcie_irq_map(const VirtBoardInfo *vbi, uint32_t gic_phandle,
757 int first_irq, const char *nodename)
758 {
759 int devfn, pin;
760 uint32_t full_irq_map[4 * 4 * 10] = { 0 };
761 uint32_t *irq_map = full_irq_map;
762
763 for (devfn = 0; devfn <= 0x18; devfn += 0x8) {
764 for (pin = 0; pin < 4; pin++) {
765 int irq_type = GIC_FDT_IRQ_TYPE_SPI;
766 int irq_nr = first_irq + ((pin + PCI_SLOT(devfn)) % PCI_NUM_PINS);
767 int irq_level = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
768 int i;
769
770 uint32_t map[] = {
771 devfn << 8, 0, 0, /* devfn */
772 pin + 1, /* PCI pin */
773 gic_phandle, 0, 0, irq_type, irq_nr, irq_level }; /* GIC irq */
774
775 /* Convert map to big endian */
776 for (i = 0; i < 10; i++) {
777 irq_map[i] = cpu_to_be32(map[i]);
778 }
779 irq_map += 10;
780 }
781 }
782
783 qemu_fdt_setprop(vbi->fdt, nodename, "interrupt-map",
784 full_irq_map, sizeof(full_irq_map));
785
786 qemu_fdt_setprop_cells(vbi->fdt, nodename, "interrupt-map-mask",
787 0x1800, 0, 0, /* devfn (PCI_SLOT(3)) */
788 0x7 /* PCI irq */);
789 }
790
791 static void create_pcie(const VirtBoardInfo *vbi, qemu_irq *pic,
792 bool use_highmem)
793 {
794 hwaddr base_mmio = vbi->memmap[VIRT_PCIE_MMIO].base;
795 hwaddr size_mmio = vbi->memmap[VIRT_PCIE_MMIO].size;
796 hwaddr base_mmio_high = vbi->memmap[VIRT_PCIE_MMIO_HIGH].base;
797 hwaddr size_mmio_high = vbi->memmap[VIRT_PCIE_MMIO_HIGH].size;
798 hwaddr base_pio = vbi->memmap[VIRT_PCIE_PIO].base;
799 hwaddr size_pio = vbi->memmap[VIRT_PCIE_PIO].size;
800 hwaddr base_ecam = vbi->memmap[VIRT_PCIE_ECAM].base;
801 hwaddr size_ecam = vbi->memmap[VIRT_PCIE_ECAM].size;
802 hwaddr base = base_mmio;
803 int nr_pcie_buses = size_ecam / PCIE_MMCFG_SIZE_MIN;
804 int irq = vbi->irqmap[VIRT_PCIE];
805 MemoryRegion *mmio_alias;
806 MemoryRegion *mmio_reg;
807 MemoryRegion *ecam_alias;
808 MemoryRegion *ecam_reg;
809 DeviceState *dev;
810 char *nodename;
811 int i;
812 PCIHostState *pci;
813
814 dev = qdev_create(NULL, TYPE_GPEX_HOST);
815 qdev_init_nofail(dev);
816
817 /* Map only the first size_ecam bytes of ECAM space */
818 ecam_alias = g_new0(MemoryRegion, 1);
819 ecam_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 0);
820 memory_region_init_alias(ecam_alias, OBJECT(dev), "pcie-ecam",
821 ecam_reg, 0, size_ecam);
822 memory_region_add_subregion(get_system_memory(), base_ecam, ecam_alias);
823
824 /* Map the MMIO window into system address space so as to expose
825 * the section of PCI MMIO space which starts at the same base address
826 * (ie 1:1 mapping for that part of PCI MMIO space visible through
827 * the window).
828 */
829 mmio_alias = g_new0(MemoryRegion, 1);
830 mmio_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 1);
831 memory_region_init_alias(mmio_alias, OBJECT(dev), "pcie-mmio",
832 mmio_reg, base_mmio, size_mmio);
833 memory_region_add_subregion(get_system_memory(), base_mmio, mmio_alias);
834
835 if (use_highmem) {
836 /* Map high MMIO space */
837 MemoryRegion *high_mmio_alias = g_new0(MemoryRegion, 1);
838
839 memory_region_init_alias(high_mmio_alias, OBJECT(dev), "pcie-mmio-high",
840 mmio_reg, base_mmio_high, size_mmio_high);
841 memory_region_add_subregion(get_system_memory(), base_mmio_high,
842 high_mmio_alias);
843 }
844
845 /* Map IO port space */
846 sysbus_mmio_map(SYS_BUS_DEVICE(dev), 2, base_pio);
847
848 for (i = 0; i < GPEX_NUM_IRQS; i++) {
849 sysbus_connect_irq(SYS_BUS_DEVICE(dev), i, pic[irq + i]);
850 }
851
852 pci = PCI_HOST_BRIDGE(dev);
853 if (pci->bus) {
854 for (i = 0; i < nb_nics; i++) {
855 NICInfo *nd = &nd_table[i];
856
857 if (!nd->model) {
858 nd->model = g_strdup("virtio");
859 }
860
861 pci_nic_init_nofail(nd, pci->bus, nd->model, NULL);
862 }
863 }
864
865 nodename = g_strdup_printf("/pcie@%" PRIx64, base);
866 qemu_fdt_add_subnode(vbi->fdt, nodename);
867 qemu_fdt_setprop_string(vbi->fdt, nodename,
868 "compatible", "pci-host-ecam-generic");
869 qemu_fdt_setprop_string(vbi->fdt, nodename, "device_type", "pci");
870 qemu_fdt_setprop_cell(vbi->fdt, nodename, "#address-cells", 3);
871 qemu_fdt_setprop_cell(vbi->fdt, nodename, "#size-cells", 2);
872 qemu_fdt_setprop_cells(vbi->fdt, nodename, "bus-range", 0,
873 nr_pcie_buses - 1);
874
875 if (vbi->v2m_phandle) {
876 qemu_fdt_setprop_cells(vbi->fdt, nodename, "msi-parent",
877 vbi->v2m_phandle);
878 }
879
880 qemu_fdt_setprop_sized_cells(vbi->fdt, nodename, "reg",
881 2, base_ecam, 2, size_ecam);
882
883 if (use_highmem) {
884 qemu_fdt_setprop_sized_cells(vbi->fdt, nodename, "ranges",
885 1, FDT_PCI_RANGE_IOPORT, 2, 0,
886 2, base_pio, 2, size_pio,
887 1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
888 2, base_mmio, 2, size_mmio,
889 1, FDT_PCI_RANGE_MMIO_64BIT,
890 2, base_mmio_high,
891 2, base_mmio_high, 2, size_mmio_high);
892 } else {
893 qemu_fdt_setprop_sized_cells(vbi->fdt, nodename, "ranges",
894 1, FDT_PCI_RANGE_IOPORT, 2, 0,
895 2, base_pio, 2, size_pio,
896 1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
897 2, base_mmio, 2, size_mmio);
898 }
899
900 qemu_fdt_setprop_cell(vbi->fdt, nodename, "#interrupt-cells", 1);
901 create_pcie_irq_map(vbi, vbi->gic_phandle, irq, nodename);
902
903 g_free(nodename);
904 }
905
906 static void create_platform_bus(VirtBoardInfo *vbi, qemu_irq *pic)
907 {
908 DeviceState *dev;
909 SysBusDevice *s;
910 int i;
911 ARMPlatformBusFDTParams *fdt_params = g_new(ARMPlatformBusFDTParams, 1);
912 MemoryRegion *sysmem = get_system_memory();
913
914 platform_bus_params.platform_bus_base = vbi->memmap[VIRT_PLATFORM_BUS].base;
915 platform_bus_params.platform_bus_size = vbi->memmap[VIRT_PLATFORM_BUS].size;
916 platform_bus_params.platform_bus_first_irq = vbi->irqmap[VIRT_PLATFORM_BUS];
917 platform_bus_params.platform_bus_num_irqs = PLATFORM_BUS_NUM_IRQS;
918
919 fdt_params->system_params = &platform_bus_params;
920 fdt_params->binfo = &vbi->bootinfo;
921 fdt_params->intc = "/intc";
922 /*
923 * register a machine init done notifier that creates the device tree
924 * nodes of the platform bus and its children dynamic sysbus devices
925 */
926 arm_register_platform_bus_fdt_creator(fdt_params);
927
928 dev = qdev_create(NULL, TYPE_PLATFORM_BUS_DEVICE);
929 dev->id = TYPE_PLATFORM_BUS_DEVICE;
930 qdev_prop_set_uint32(dev, "num_irqs",
931 platform_bus_params.platform_bus_num_irqs);
932 qdev_prop_set_uint32(dev, "mmio_size",
933 platform_bus_params.platform_bus_size);
934 qdev_init_nofail(dev);
935 s = SYS_BUS_DEVICE(dev);
936
937 for (i = 0; i < platform_bus_params.platform_bus_num_irqs; i++) {
938 int irqn = platform_bus_params.platform_bus_first_irq + i;
939 sysbus_connect_irq(s, i, pic[irqn]);
940 }
941
942 memory_region_add_subregion(sysmem,
943 platform_bus_params.platform_bus_base,
944 sysbus_mmio_get_region(s, 0));
945 }
946
947 static void *machvirt_dtb(const struct arm_boot_info *binfo, int *fdt_size)
948 {
949 const VirtBoardInfo *board = (const VirtBoardInfo *)binfo;
950
951 *fdt_size = board->fdt_size;
952 return board->fdt;
953 }
954
955 static void virt_build_smbios(VirtGuestInfo *guest_info)
956 {
957 FWCfgState *fw_cfg = guest_info->fw_cfg;
958 uint8_t *smbios_tables, *smbios_anchor;
959 size_t smbios_tables_len, smbios_anchor_len;
960 const char *product = "QEMU Virtual Machine";
961
962 if (!fw_cfg) {
963 return;
964 }
965
966 if (kvm_enabled()) {
967 product = "KVM Virtual Machine";
968 }
969
970 smbios_set_defaults("QEMU", product,
971 "1.0", false, true, SMBIOS_ENTRY_POINT_30);
972
973 smbios_get_tables(NULL, 0, &smbios_tables, &smbios_tables_len,
974 &smbios_anchor, &smbios_anchor_len);
975
976 if (smbios_anchor) {
977 fw_cfg_add_file(fw_cfg, "etc/smbios/smbios-tables",
978 smbios_tables, smbios_tables_len);
979 fw_cfg_add_file(fw_cfg, "etc/smbios/smbios-anchor",
980 smbios_anchor, smbios_anchor_len);
981 }
982 }
983
984 static
985 void virt_guest_info_machine_done(Notifier *notifier, void *data)
986 {
987 VirtGuestInfoState *guest_info_state = container_of(notifier,
988 VirtGuestInfoState, machine_done);
989 virt_acpi_setup(&guest_info_state->info);
990 virt_build_smbios(&guest_info_state->info);
991 }
992
993 static void machvirt_init(MachineState *machine)
994 {
995 VirtMachineState *vms = VIRT_MACHINE(machine);
996 qemu_irq pic[NUM_IRQS];
997 MemoryRegion *sysmem = get_system_memory();
998 int gic_version = vms->gic_version;
999 int n, max_cpus;
1000 MemoryRegion *ram = g_new(MemoryRegion, 1);
1001 const char *cpu_model = machine->cpu_model;
1002 VirtBoardInfo *vbi;
1003 VirtGuestInfoState *guest_info_state = g_malloc0(sizeof *guest_info_state);
1004 VirtGuestInfo *guest_info = &guest_info_state->info;
1005 char **cpustr;
1006
1007 if (!cpu_model) {
1008 cpu_model = "cortex-a15";
1009 }
1010
1011 /* We can probe only here because during property set
1012 * KVM is not available yet
1013 */
1014 if (!gic_version) {
1015 gic_version = kvm_arm_vgic_probe();
1016 if (!gic_version) {
1017 error_report("Unable to determine GIC version supported by host");
1018 error_printf("KVM acceleration is probably not supported\n");
1019 exit(1);
1020 }
1021 }
1022
1023 /* Separate the actual CPU model name from any appended features */
1024 cpustr = g_strsplit(cpu_model, ",", 2);
1025
1026 vbi = find_machine_info(cpustr[0]);
1027
1028 if (!vbi) {
1029 error_report("mach-virt: CPU %s not supported", cpustr[0]);
1030 exit(1);
1031 }
1032
1033 /* The maximum number of CPUs depends on the GIC version, or on how
1034 * many redistributors we can fit into the memory map.
1035 */
1036 if (gic_version == 3) {
1037 max_cpus = vbi->memmap[VIRT_GIC_REDIST].size / 0x20000;
1038 } else {
1039 max_cpus = GIC_NCPU;
1040 }
1041
1042 if (smp_cpus > max_cpus) {
1043 error_report("Number of SMP CPUs requested (%d) exceeds max CPUs "
1044 "supported by machine 'mach-virt' (%d)",
1045 smp_cpus, max_cpus);
1046 exit(1);
1047 }
1048
1049 vbi->smp_cpus = smp_cpus;
1050
1051 if (machine->ram_size > vbi->memmap[VIRT_MEM].size) {
1052 error_report("mach-virt: cannot model more than 30GB RAM");
1053 exit(1);
1054 }
1055
1056 create_fdt(vbi);
1057
1058 for (n = 0; n < smp_cpus; n++) {
1059 ObjectClass *oc = cpu_class_by_name(TYPE_ARM_CPU, cpustr[0]);
1060 CPUClass *cc = CPU_CLASS(oc);
1061 Object *cpuobj;
1062 Error *err = NULL;
1063 char *cpuopts = g_strdup(cpustr[1]);
1064
1065 if (!oc) {
1066 error_report("Unable to find CPU definition");
1067 exit(1);
1068 }
1069 cpuobj = object_new(object_class_get_name(oc));
1070
1071 /* Handle any CPU options specified by the user */
1072 cc->parse_features(CPU(cpuobj), cpuopts, &err);
1073 g_free(cpuopts);
1074 if (err) {
1075 error_report_err(err);
1076 exit(1);
1077 }
1078
1079 if (!vms->secure) {
1080 object_property_set_bool(cpuobj, false, "has_el3", NULL);
1081 }
1082
1083 object_property_set_int(cpuobj, QEMU_PSCI_CONDUIT_HVC, "psci-conduit",
1084 NULL);
1085
1086 /* Secondary CPUs start in PSCI powered-down state */
1087 if (n > 0) {
1088 object_property_set_bool(cpuobj, true, "start-powered-off", NULL);
1089 }
1090
1091 if (object_property_find(cpuobj, "reset-cbar", NULL)) {
1092 object_property_set_int(cpuobj, vbi->memmap[VIRT_CPUPERIPHS].base,
1093 "reset-cbar", &error_abort);
1094 }
1095
1096 object_property_set_bool(cpuobj, true, "realized", NULL);
1097 }
1098 g_strfreev(cpustr);
1099 fdt_add_timer_nodes(vbi, gic_version);
1100 fdt_add_cpu_nodes(vbi);
1101 fdt_add_psci_node(vbi);
1102
1103 memory_region_allocate_system_memory(ram, NULL, "mach-virt.ram",
1104 machine->ram_size);
1105 memory_region_add_subregion(sysmem, vbi->memmap[VIRT_MEM].base, ram);
1106
1107 create_flash(vbi);
1108
1109 create_gic(vbi, pic, gic_version, vms->secure);
1110
1111 create_uart(vbi, pic);
1112
1113 create_rtc(vbi, pic);
1114
1115 create_pcie(vbi, pic, vms->highmem);
1116
1117 create_gpio(vbi, pic);
1118
1119 /* Create mmio transports, so the user can create virtio backends
1120 * (which will be automatically plugged in to the transports). If
1121 * no backend is created the transport will just sit harmlessly idle.
1122 */
1123 create_virtio_devices(vbi, pic);
1124
1125 create_fw_cfg(vbi, &address_space_memory);
1126 rom_set_fw(fw_cfg_find());
1127
1128 guest_info->smp_cpus = smp_cpus;
1129 guest_info->fw_cfg = fw_cfg_find();
1130 guest_info->memmap = vbi->memmap;
1131 guest_info->irqmap = vbi->irqmap;
1132 guest_info->use_highmem = vms->highmem;
1133 guest_info->gic_version = gic_version;
1134 guest_info_state->machine_done.notify = virt_guest_info_machine_done;
1135 qemu_add_machine_init_done_notifier(&guest_info_state->machine_done);
1136
1137 vbi->bootinfo.ram_size = machine->ram_size;
1138 vbi->bootinfo.kernel_filename = machine->kernel_filename;
1139 vbi->bootinfo.kernel_cmdline = machine->kernel_cmdline;
1140 vbi->bootinfo.initrd_filename = machine->initrd_filename;
1141 vbi->bootinfo.nb_cpus = smp_cpus;
1142 vbi->bootinfo.board_id = -1;
1143 vbi->bootinfo.loader_start = vbi->memmap[VIRT_MEM].base;
1144 vbi->bootinfo.get_dtb = machvirt_dtb;
1145 vbi->bootinfo.firmware_loaded = bios_name || drive_get(IF_PFLASH, 0, 0);
1146 arm_load_kernel(ARM_CPU(first_cpu), &vbi->bootinfo);
1147
1148 /*
1149 * arm_load_kernel machine init done notifier registration must
1150 * happen before the platform_bus_create call. In this latter,
1151 * another notifier is registered which adds platform bus nodes.
1152 * Notifiers are executed in registration reverse order.
1153 */
1154 create_platform_bus(vbi, pic);
1155 }
1156
1157 static bool virt_get_secure(Object *obj, Error **errp)
1158 {
1159 VirtMachineState *vms = VIRT_MACHINE(obj);
1160
1161 return vms->secure;
1162 }
1163
1164 static void virt_set_secure(Object *obj, bool value, Error **errp)
1165 {
1166 VirtMachineState *vms = VIRT_MACHINE(obj);
1167
1168 vms->secure = value;
1169 }
1170
1171 static bool virt_get_highmem(Object *obj, Error **errp)
1172 {
1173 VirtMachineState *vms = VIRT_MACHINE(obj);
1174
1175 return vms->highmem;
1176 }
1177
1178 static void virt_set_highmem(Object *obj, bool value, Error **errp)
1179 {
1180 VirtMachineState *vms = VIRT_MACHINE(obj);
1181
1182 vms->highmem = value;
1183 }
1184
1185 static char *virt_get_gic_version(Object *obj, Error **errp)
1186 {
1187 VirtMachineState *vms = VIRT_MACHINE(obj);
1188 const char *val = vms->gic_version == 3 ? "3" : "2";
1189
1190 return g_strdup(val);
1191 }
1192
1193 static void virt_set_gic_version(Object *obj, const char *value, Error **errp)
1194 {
1195 VirtMachineState *vms = VIRT_MACHINE(obj);
1196
1197 if (!strcmp(value, "3")) {
1198 vms->gic_version = 3;
1199 } else if (!strcmp(value, "2")) {
1200 vms->gic_version = 2;
1201 } else if (!strcmp(value, "host")) {
1202 vms->gic_version = 0; /* Will probe later */
1203 } else {
1204 error_setg(errp, "Invalid gic-version value");
1205 error_append_hint(errp, "Valid values are 3, 2, host.\n");
1206 }
1207 }
1208
1209 static void virt_instance_init(Object *obj)
1210 {
1211 VirtMachineState *vms = VIRT_MACHINE(obj);
1212
1213 /* EL3 is disabled by default on virt: this makes us consistent
1214 * between KVM and TCG for this board, and it also allows us to
1215 * boot UEFI blobs which assume no TrustZone support.
1216 */
1217 vms->secure = false;
1218 object_property_add_bool(obj, "secure", virt_get_secure,
1219 virt_set_secure, NULL);
1220 object_property_set_description(obj, "secure",
1221 "Set on/off to enable/disable the ARM "
1222 "Security Extensions (TrustZone)",
1223 NULL);
1224
1225 /* High memory is enabled by default */
1226 vms->highmem = true;
1227 object_property_add_bool(obj, "highmem", virt_get_highmem,
1228 virt_set_highmem, NULL);
1229 object_property_set_description(obj, "highmem",
1230 "Set on/off to enable/disable using "
1231 "physical address space above 32 bits",
1232 NULL);
1233 /* Default GIC type is v2 */
1234 vms->gic_version = 2;
1235 object_property_add_str(obj, "gic-version", virt_get_gic_version,
1236 virt_set_gic_version, NULL);
1237 object_property_set_description(obj, "gic-version",
1238 "Set GIC version. "
1239 "Valid values are 2, 3 and host", NULL);
1240 }
1241
1242 static void virt_class_init(ObjectClass *oc, void *data)
1243 {
1244 MachineClass *mc = MACHINE_CLASS(oc);
1245
1246 mc->desc = "ARM Virtual Machine",
1247 mc->init = machvirt_init;
1248 /* Start max_cpus at the maximum QEMU supports. We'll further restrict
1249 * it later in machvirt_init, where we have more information about the
1250 * configuration of the particular instance.
1251 */
1252 mc->max_cpus = MAX_CPUMASK_BITS;
1253 mc->has_dynamic_sysbus = true;
1254 mc->block_default_type = IF_VIRTIO;
1255 mc->no_cdrom = 1;
1256 mc->pci_allow_0_address = true;
1257 }
1258
1259 static const TypeInfo machvirt_info = {
1260 .name = TYPE_VIRT_MACHINE,
1261 .parent = TYPE_MACHINE,
1262 .instance_size = sizeof(VirtMachineState),
1263 .instance_init = virt_instance_init,
1264 .class_size = sizeof(VirtMachineClass),
1265 .class_init = virt_class_init,
1266 };
1267
1268 static void machvirt_machine_init(void)
1269 {
1270 type_register_static(&machvirt_info);
1271 }
1272
1273 machine_init(machvirt_machine_init);