xref: /openbmc/qemu/hw/arm/virt.c (revision ac1d8878)
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 "hw/sysbus.h"
32 #include "hw/arm/arm.h"
33 #include "hw/arm/primecell.h"
34 #include "hw/arm/virt.h"
35 #include "hw/devices.h"
36 #include "net/net.h"
37 #include "sysemu/block-backend.h"
38 #include "sysemu/device_tree.h"
39 #include "sysemu/sysemu.h"
40 #include "sysemu/kvm.h"
41 #include "hw/boards.h"
42 #include "hw/loader.h"
43 #include "exec/address-spaces.h"
44 #include "qemu/bitops.h"
45 #include "qemu/error-report.h"
46 #include "hw/pci-host/gpex.h"
47 #include "hw/arm/virt-acpi-build.h"
48 #include "hw/arm/sysbus-fdt.h"
49 #include "hw/platform-bus.h"
50 #include "hw/arm/fdt.h"
51 #include "hw/intc/arm_gic_common.h"
52 #include "kvm_arm.h"
53 #include "hw/smbios/smbios.h"
54 #include "qapi/visitor.h"
55 #include "standard-headers/linux/input.h"
56 
57 /* Number of external interrupt lines to configure the GIC with */
58 #define NUM_IRQS 256
59 
60 #define PLATFORM_BUS_NUM_IRQS 64
61 
62 static ARMPlatformBusSystemParams platform_bus_params;
63 
64 typedef struct VirtBoardInfo {
65     struct arm_boot_info bootinfo;
66     const char *cpu_model;
67     const MemMapEntry *memmap;
68     const int *irqmap;
69     int smp_cpus;
70     void *fdt;
71     int fdt_size;
72     uint32_t clock_phandle;
73     uint32_t gic_phandle;
74     uint32_t v2m_phandle;
75 } VirtBoardInfo;
76 
77 typedef struct {
78     MachineClass parent;
79     VirtBoardInfo *daughterboard;
80 } VirtMachineClass;
81 
82 typedef struct {
83     MachineState parent;
84     bool secure;
85     bool highmem;
86     int32_t gic_version;
87 } VirtMachineState;
88 
89 #define TYPE_VIRT_MACHINE   MACHINE_TYPE_NAME("virt")
90 #define VIRT_MACHINE(obj) \
91     OBJECT_CHECK(VirtMachineState, (obj), TYPE_VIRT_MACHINE)
92 #define VIRT_MACHINE_GET_CLASS(obj) \
93     OBJECT_GET_CLASS(VirtMachineClass, obj, TYPE_VIRT_MACHINE)
94 #define VIRT_MACHINE_CLASS(klass) \
95     OBJECT_CLASS_CHECK(VirtMachineClass, klass, TYPE_VIRT_MACHINE)
96 
97 /* Addresses and sizes of our components.
98  * 0..128MB is space for a flash device so we can run bootrom code such as UEFI.
99  * 128MB..256MB is used for miscellaneous device I/O.
100  * 256MB..1GB is reserved for possible future PCI support (ie where the
101  * PCI memory window will go if we add a PCI host controller).
102  * 1GB and up is RAM (which may happily spill over into the
103  * high memory region beyond 4GB).
104  * This represents a compromise between how much RAM can be given to
105  * a 32 bit VM and leaving space for expansion and in particular for PCI.
106  * Note that devices should generally be placed at multiples of 0x10000,
107  * to accommodate guests using 64K pages.
108  */
109 static const MemMapEntry a15memmap[] = {
110     /* Space up to 0x8000000 is reserved for a boot ROM */
111     [VIRT_FLASH] =              {          0, 0x08000000 },
112     [VIRT_CPUPERIPHS] =         { 0x08000000, 0x00020000 },
113     /* GIC distributor and CPU interfaces sit inside the CPU peripheral space */
114     [VIRT_GIC_DIST] =           { 0x08000000, 0x00010000 },
115     [VIRT_GIC_CPU] =            { 0x08010000, 0x00010000 },
116     [VIRT_GIC_V2M] =            { 0x08020000, 0x00001000 },
117     /* The space in between here is reserved for GICv3 CPU/vCPU/HYP */
118     [VIRT_GIC_ITS] =            { 0x08080000, 0x00020000 },
119     /* This redistributor space allows up to 2*64kB*123 CPUs */
120     [VIRT_GIC_REDIST] =         { 0x080A0000, 0x00F60000 },
121     [VIRT_UART] =               { 0x09000000, 0x00001000 },
122     [VIRT_RTC] =                { 0x09010000, 0x00001000 },
123     [VIRT_FW_CFG] =             { 0x09020000, 0x00000018 },
124     [VIRT_GPIO] =               { 0x09030000, 0x00001000 },
125     [VIRT_MMIO] =               { 0x0a000000, 0x00000200 },
126     /* ...repeating for a total of NUM_VIRTIO_TRANSPORTS, each of that size */
127     [VIRT_PLATFORM_BUS] =       { 0x0c000000, 0x02000000 },
128     [VIRT_PCIE_MMIO] =          { 0x10000000, 0x2eff0000 },
129     [VIRT_PCIE_PIO] =           { 0x3eff0000, 0x00010000 },
130     [VIRT_PCIE_ECAM] =          { 0x3f000000, 0x01000000 },
131     [VIRT_MEM] =                { 0x40000000, 30ULL * 1024 * 1024 * 1024 },
132     /* Second PCIe window, 512GB wide at the 512GB boundary */
133     [VIRT_PCIE_MMIO_HIGH] =   { 0x8000000000ULL, 0x8000000000ULL },
134 };
135 
136 static const int a15irqmap[] = {
137     [VIRT_UART] = 1,
138     [VIRT_RTC] = 2,
139     [VIRT_PCIE] = 3, /* ... to 6 */
140     [VIRT_GPIO] = 7,
141     [VIRT_MMIO] = 16, /* ...to 16 + NUM_VIRTIO_TRANSPORTS - 1 */
142     [VIRT_GIC_V2M] = 48, /* ...to 48 + NUM_GICV2M_SPIS - 1 */
143     [VIRT_PLATFORM_BUS] = 112, /* ...to 112 + PLATFORM_BUS_NUM_IRQS -1 */
144 };
145 
146 static VirtBoardInfo machines[] = {
147     {
148         .cpu_model = "cortex-a15",
149         .memmap = a15memmap,
150         .irqmap = a15irqmap,
151     },
152     {
153         .cpu_model = "cortex-a53",
154         .memmap = a15memmap,
155         .irqmap = a15irqmap,
156     },
157     {
158         .cpu_model = "cortex-a57",
159         .memmap = a15memmap,
160         .irqmap = a15irqmap,
161     },
162     {
163         .cpu_model = "host",
164         .memmap = a15memmap,
165         .irqmap = a15irqmap,
166     },
167 };
168 
169 static VirtBoardInfo *find_machine_info(const char *cpu)
170 {
171     int i;
172 
173     for (i = 0; i < ARRAY_SIZE(machines); i++) {
174         if (strcmp(cpu, machines[i].cpu_model) == 0) {
175             return &machines[i];
176         }
177     }
178     return NULL;
179 }
180 
181 static void create_fdt(VirtBoardInfo *vbi)
182 {
183     void *fdt = create_device_tree(&vbi->fdt_size);
184 
185     if (!fdt) {
186         error_report("create_device_tree() failed");
187         exit(1);
188     }
189 
190     vbi->fdt = fdt;
191 
192     /* Header */
193     qemu_fdt_setprop_string(fdt, "/", "compatible", "linux,dummy-virt");
194     qemu_fdt_setprop_cell(fdt, "/", "#address-cells", 0x2);
195     qemu_fdt_setprop_cell(fdt, "/", "#size-cells", 0x2);
196 
197     /*
198      * /chosen and /memory nodes must exist for load_dtb
199      * to fill in necessary properties later
200      */
201     qemu_fdt_add_subnode(fdt, "/chosen");
202     qemu_fdt_add_subnode(fdt, "/memory");
203     qemu_fdt_setprop_string(fdt, "/memory", "device_type", "memory");
204 
205     /* Clock node, for the benefit of the UART. The kernel device tree
206      * binding documentation claims the PL011 node clock properties are
207      * optional but in practice if you omit them the kernel refuses to
208      * probe for the device.
209      */
210     vbi->clock_phandle = qemu_fdt_alloc_phandle(fdt);
211     qemu_fdt_add_subnode(fdt, "/apb-pclk");
212     qemu_fdt_setprop_string(fdt, "/apb-pclk", "compatible", "fixed-clock");
213     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "#clock-cells", 0x0);
214     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "clock-frequency", 24000000);
215     qemu_fdt_setprop_string(fdt, "/apb-pclk", "clock-output-names",
216                                 "clk24mhz");
217     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "phandle", vbi->clock_phandle);
218 
219 }
220 
221 static void fdt_add_psci_node(const VirtBoardInfo *vbi)
222 {
223     uint32_t cpu_suspend_fn;
224     uint32_t cpu_off_fn;
225     uint32_t cpu_on_fn;
226     uint32_t migrate_fn;
227     void *fdt = vbi->fdt;
228     ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(0));
229 
230     qemu_fdt_add_subnode(fdt, "/psci");
231     if (armcpu->psci_version == 2) {
232         const char comp[] = "arm,psci-0.2\0arm,psci";
233         qemu_fdt_setprop(fdt, "/psci", "compatible", comp, sizeof(comp));
234 
235         cpu_off_fn = QEMU_PSCI_0_2_FN_CPU_OFF;
236         if (arm_feature(&armcpu->env, ARM_FEATURE_AARCH64)) {
237             cpu_suspend_fn = QEMU_PSCI_0_2_FN64_CPU_SUSPEND;
238             cpu_on_fn = QEMU_PSCI_0_2_FN64_CPU_ON;
239             migrate_fn = QEMU_PSCI_0_2_FN64_MIGRATE;
240         } else {
241             cpu_suspend_fn = QEMU_PSCI_0_2_FN_CPU_SUSPEND;
242             cpu_on_fn = QEMU_PSCI_0_2_FN_CPU_ON;
243             migrate_fn = QEMU_PSCI_0_2_FN_MIGRATE;
244         }
245     } else {
246         qemu_fdt_setprop_string(fdt, "/psci", "compatible", "arm,psci");
247 
248         cpu_suspend_fn = QEMU_PSCI_0_1_FN_CPU_SUSPEND;
249         cpu_off_fn = QEMU_PSCI_0_1_FN_CPU_OFF;
250         cpu_on_fn = QEMU_PSCI_0_1_FN_CPU_ON;
251         migrate_fn = QEMU_PSCI_0_1_FN_MIGRATE;
252     }
253 
254     /* We adopt the PSCI spec's nomenclature, and use 'conduit' to refer
255      * to the instruction that should be used to invoke PSCI functions.
256      * However, the device tree binding uses 'method' instead, so that is
257      * what we should use here.
258      */
259     qemu_fdt_setprop_string(fdt, "/psci", "method", "hvc");
260 
261     qemu_fdt_setprop_cell(fdt, "/psci", "cpu_suspend", cpu_suspend_fn);
262     qemu_fdt_setprop_cell(fdt, "/psci", "cpu_off", cpu_off_fn);
263     qemu_fdt_setprop_cell(fdt, "/psci", "cpu_on", cpu_on_fn);
264     qemu_fdt_setprop_cell(fdt, "/psci", "migrate", migrate_fn);
265 }
266 
267 static void fdt_add_timer_nodes(const VirtBoardInfo *vbi, int gictype)
268 {
269     /* Note that on A15 h/w these interrupts are level-triggered,
270      * but for the GIC implementation provided by both QEMU and KVM
271      * they are edge-triggered.
272      */
273     ARMCPU *armcpu;
274     uint32_t irqflags = GIC_FDT_IRQ_FLAGS_EDGE_LO_HI;
275 
276     if (gictype == 2) {
277         irqflags = deposit32(irqflags, GIC_FDT_IRQ_PPI_CPU_START,
278                              GIC_FDT_IRQ_PPI_CPU_WIDTH,
279                              (1 << vbi->smp_cpus) - 1);
280     }
281 
282     qemu_fdt_add_subnode(vbi->fdt, "/timer");
283 
284     armcpu = ARM_CPU(qemu_get_cpu(0));
285     if (arm_feature(&armcpu->env, ARM_FEATURE_V8)) {
286         const char compat[] = "arm,armv8-timer\0arm,armv7-timer";
287         qemu_fdt_setprop(vbi->fdt, "/timer", "compatible",
288                          compat, sizeof(compat));
289     } else {
290         qemu_fdt_setprop_string(vbi->fdt, "/timer", "compatible",
291                                 "arm,armv7-timer");
292     }
293     qemu_fdt_setprop_cells(vbi->fdt, "/timer", "interrupts",
294                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_S_EL1_IRQ, irqflags,
295                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_NS_EL1_IRQ, irqflags,
296                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_VIRT_IRQ, irqflags,
297                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_NS_EL2_IRQ, irqflags);
298 }
299 
300 static void fdt_add_cpu_nodes(const VirtBoardInfo *vbi)
301 {
302     int cpu;
303     int addr_cells = 1;
304 
305     /*
306      * From Documentation/devicetree/bindings/arm/cpus.txt
307      *  On ARM v8 64-bit systems value should be set to 2,
308      *  that corresponds to the MPIDR_EL1 register size.
309      *  If MPIDR_EL1[63:32] value is equal to 0 on all CPUs
310      *  in the system, #address-cells can be set to 1, since
311      *  MPIDR_EL1[63:32] bits are not used for CPUs
312      *  identification.
313      *
314      *  Here we actually don't know whether our system is 32- or 64-bit one.
315      *  The simplest way to go is to examine affinity IDs of all our CPUs. If
316      *  at least one of them has Aff3 populated, we set #address-cells to 2.
317      */
318     for (cpu = 0; cpu < vbi->smp_cpus; cpu++) {
319         ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
320 
321         if (armcpu->mp_affinity & ARM_AFF3_MASK) {
322             addr_cells = 2;
323             break;
324         }
325     }
326 
327     qemu_fdt_add_subnode(vbi->fdt, "/cpus");
328     qemu_fdt_setprop_cell(vbi->fdt, "/cpus", "#address-cells", addr_cells);
329     qemu_fdt_setprop_cell(vbi->fdt, "/cpus", "#size-cells", 0x0);
330 
331     for (cpu = vbi->smp_cpus - 1; cpu >= 0; cpu--) {
332         char *nodename = g_strdup_printf("/cpus/cpu@%d", cpu);
333         ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
334 
335         qemu_fdt_add_subnode(vbi->fdt, nodename);
336         qemu_fdt_setprop_string(vbi->fdt, nodename, "device_type", "cpu");
337         qemu_fdt_setprop_string(vbi->fdt, nodename, "compatible",
338                                     armcpu->dtb_compatible);
339 
340         if (vbi->smp_cpus > 1) {
341             qemu_fdt_setprop_string(vbi->fdt, nodename,
342                                         "enable-method", "psci");
343         }
344 
345         if (addr_cells == 2) {
346             qemu_fdt_setprop_u64(vbi->fdt, nodename, "reg",
347                                  armcpu->mp_affinity);
348         } else {
349             qemu_fdt_setprop_cell(vbi->fdt, nodename, "reg",
350                                   armcpu->mp_affinity);
351         }
352 
353         g_free(nodename);
354     }
355 }
356 
357 static void fdt_add_v2m_gic_node(VirtBoardInfo *vbi)
358 {
359     vbi->v2m_phandle = qemu_fdt_alloc_phandle(vbi->fdt);
360     qemu_fdt_add_subnode(vbi->fdt, "/intc/v2m");
361     qemu_fdt_setprop_string(vbi->fdt, "/intc/v2m", "compatible",
362                             "arm,gic-v2m-frame");
363     qemu_fdt_setprop(vbi->fdt, "/intc/v2m", "msi-controller", NULL, 0);
364     qemu_fdt_setprop_sized_cells(vbi->fdt, "/intc/v2m", "reg",
365                                  2, vbi->memmap[VIRT_GIC_V2M].base,
366                                  2, vbi->memmap[VIRT_GIC_V2M].size);
367     qemu_fdt_setprop_cell(vbi->fdt, "/intc/v2m", "phandle", vbi->v2m_phandle);
368 }
369 
370 static void fdt_add_gic_node(VirtBoardInfo *vbi, int type)
371 {
372     vbi->gic_phandle = qemu_fdt_alloc_phandle(vbi->fdt);
373     qemu_fdt_setprop_cell(vbi->fdt, "/", "interrupt-parent", vbi->gic_phandle);
374 
375     qemu_fdt_add_subnode(vbi->fdt, "/intc");
376     qemu_fdt_setprop_cell(vbi->fdt, "/intc", "#interrupt-cells", 3);
377     qemu_fdt_setprop(vbi->fdt, "/intc", "interrupt-controller", NULL, 0);
378     qemu_fdt_setprop_cell(vbi->fdt, "/intc", "#address-cells", 0x2);
379     qemu_fdt_setprop_cell(vbi->fdt, "/intc", "#size-cells", 0x2);
380     qemu_fdt_setprop(vbi->fdt, "/intc", "ranges", NULL, 0);
381     if (type == 3) {
382         qemu_fdt_setprop_string(vbi->fdt, "/intc", "compatible",
383                                 "arm,gic-v3");
384         qemu_fdt_setprop_sized_cells(vbi->fdt, "/intc", "reg",
385                                      2, vbi->memmap[VIRT_GIC_DIST].base,
386                                      2, vbi->memmap[VIRT_GIC_DIST].size,
387                                      2, vbi->memmap[VIRT_GIC_REDIST].base,
388                                      2, vbi->memmap[VIRT_GIC_REDIST].size);
389     } else {
390         /* 'cortex-a15-gic' means 'GIC v2' */
391         qemu_fdt_setprop_string(vbi->fdt, "/intc", "compatible",
392                                 "arm,cortex-a15-gic");
393         qemu_fdt_setprop_sized_cells(vbi->fdt, "/intc", "reg",
394                                       2, vbi->memmap[VIRT_GIC_DIST].base,
395                                       2, vbi->memmap[VIRT_GIC_DIST].size,
396                                       2, vbi->memmap[VIRT_GIC_CPU].base,
397                                       2, vbi->memmap[VIRT_GIC_CPU].size);
398     }
399 
400     qemu_fdt_setprop_cell(vbi->fdt, "/intc", "phandle", vbi->gic_phandle);
401 }
402 
403 static void create_v2m(VirtBoardInfo *vbi, qemu_irq *pic)
404 {
405     int i;
406     int irq = vbi->irqmap[VIRT_GIC_V2M];
407     DeviceState *dev;
408 
409     dev = qdev_create(NULL, "arm-gicv2m");
410     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vbi->memmap[VIRT_GIC_V2M].base);
411     qdev_prop_set_uint32(dev, "base-spi", irq);
412     qdev_prop_set_uint32(dev, "num-spi", NUM_GICV2M_SPIS);
413     qdev_init_nofail(dev);
414 
415     for (i = 0; i < NUM_GICV2M_SPIS; i++) {
416         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i, pic[irq + i]);
417     }
418 
419     fdt_add_v2m_gic_node(vbi);
420 }
421 
422 static void create_gic(VirtBoardInfo *vbi, qemu_irq *pic, int type, bool secure)
423 {
424     /* We create a standalone GIC */
425     DeviceState *gicdev;
426     SysBusDevice *gicbusdev;
427     const char *gictype;
428     int i;
429 
430     gictype = (type == 3) ? gicv3_class_name() : gic_class_name();
431 
432     gicdev = qdev_create(NULL, gictype);
433     qdev_prop_set_uint32(gicdev, "revision", type);
434     qdev_prop_set_uint32(gicdev, "num-cpu", smp_cpus);
435     /* Note that the num-irq property counts both internal and external
436      * interrupts; there are always 32 of the former (mandated by GIC spec).
437      */
438     qdev_prop_set_uint32(gicdev, "num-irq", NUM_IRQS + 32);
439     if (!kvm_irqchip_in_kernel()) {
440         qdev_prop_set_bit(gicdev, "has-security-extensions", secure);
441     }
442     qdev_init_nofail(gicdev);
443     gicbusdev = SYS_BUS_DEVICE(gicdev);
444     sysbus_mmio_map(gicbusdev, 0, vbi->memmap[VIRT_GIC_DIST].base);
445     if (type == 3) {
446         sysbus_mmio_map(gicbusdev, 1, vbi->memmap[VIRT_GIC_REDIST].base);
447     } else {
448         sysbus_mmio_map(gicbusdev, 1, vbi->memmap[VIRT_GIC_CPU].base);
449     }
450 
451     /* Wire the outputs from each CPU's generic timer to the
452      * appropriate GIC PPI inputs, and the GIC's IRQ output to
453      * the CPU's IRQ input.
454      */
455     for (i = 0; i < smp_cpus; i++) {
456         DeviceState *cpudev = DEVICE(qemu_get_cpu(i));
457         int ppibase = NUM_IRQS + i * GIC_INTERNAL + GIC_NR_SGIS;
458         int irq;
459         /* Mapping from the output timer irq lines from the CPU to the
460          * GIC PPI inputs we use for the virt board.
461          */
462         const int timer_irq[] = {
463             [GTIMER_PHYS] = ARCH_TIMER_NS_EL1_IRQ,
464             [GTIMER_VIRT] = ARCH_TIMER_VIRT_IRQ,
465             [GTIMER_HYP]  = ARCH_TIMER_NS_EL2_IRQ,
466             [GTIMER_SEC]  = ARCH_TIMER_S_EL1_IRQ,
467         };
468 
469         for (irq = 0; irq < ARRAY_SIZE(timer_irq); irq++) {
470             qdev_connect_gpio_out(cpudev, irq,
471                                   qdev_get_gpio_in(gicdev,
472                                                    ppibase + timer_irq[irq]));
473         }
474 
475         sysbus_connect_irq(gicbusdev, i, qdev_get_gpio_in(cpudev, ARM_CPU_IRQ));
476         sysbus_connect_irq(gicbusdev, i + smp_cpus,
477                            qdev_get_gpio_in(cpudev, ARM_CPU_FIQ));
478     }
479 
480     for (i = 0; i < NUM_IRQS; i++) {
481         pic[i] = qdev_get_gpio_in(gicdev, i);
482     }
483 
484     fdt_add_gic_node(vbi, type);
485 
486     if (type == 2) {
487         create_v2m(vbi, pic);
488     }
489 }
490 
491 static void create_uart(const VirtBoardInfo *vbi, qemu_irq *pic)
492 {
493     char *nodename;
494     hwaddr base = vbi->memmap[VIRT_UART].base;
495     hwaddr size = vbi->memmap[VIRT_UART].size;
496     int irq = vbi->irqmap[VIRT_UART];
497     const char compat[] = "arm,pl011\0arm,primecell";
498     const char clocknames[] = "uartclk\0apb_pclk";
499 
500     sysbus_create_simple("pl011", base, pic[irq]);
501 
502     nodename = g_strdup_printf("/pl011@%" PRIx64, base);
503     qemu_fdt_add_subnode(vbi->fdt, nodename);
504     /* Note that we can't use setprop_string because of the embedded NUL */
505     qemu_fdt_setprop(vbi->fdt, nodename, "compatible",
506                          compat, sizeof(compat));
507     qemu_fdt_setprop_sized_cells(vbi->fdt, nodename, "reg",
508                                      2, base, 2, size);
509     qemu_fdt_setprop_cells(vbi->fdt, nodename, "interrupts",
510                                GIC_FDT_IRQ_TYPE_SPI, irq,
511                                GIC_FDT_IRQ_FLAGS_LEVEL_HI);
512     qemu_fdt_setprop_cells(vbi->fdt, nodename, "clocks",
513                                vbi->clock_phandle, vbi->clock_phandle);
514     qemu_fdt_setprop(vbi->fdt, nodename, "clock-names",
515                          clocknames, sizeof(clocknames));
516 
517     qemu_fdt_setprop_string(vbi->fdt, "/chosen", "stdout-path", nodename);
518     g_free(nodename);
519 }
520 
521 static void create_rtc(const VirtBoardInfo *vbi, qemu_irq *pic)
522 {
523     char *nodename;
524     hwaddr base = vbi->memmap[VIRT_RTC].base;
525     hwaddr size = vbi->memmap[VIRT_RTC].size;
526     int irq = vbi->irqmap[VIRT_RTC];
527     const char compat[] = "arm,pl031\0arm,primecell";
528 
529     sysbus_create_simple("pl031", base, pic[irq]);
530 
531     nodename = g_strdup_printf("/pl031@%" PRIx64, base);
532     qemu_fdt_add_subnode(vbi->fdt, nodename);
533     qemu_fdt_setprop(vbi->fdt, nodename, "compatible", compat, sizeof(compat));
534     qemu_fdt_setprop_sized_cells(vbi->fdt, nodename, "reg",
535                                  2, base, 2, size);
536     qemu_fdt_setprop_cells(vbi->fdt, nodename, "interrupts",
537                            GIC_FDT_IRQ_TYPE_SPI, irq,
538                            GIC_FDT_IRQ_FLAGS_LEVEL_HI);
539     qemu_fdt_setprop_cell(vbi->fdt, nodename, "clocks", vbi->clock_phandle);
540     qemu_fdt_setprop_string(vbi->fdt, nodename, "clock-names", "apb_pclk");
541     g_free(nodename);
542 }
543 
544 static DeviceState *pl061_dev;
545 static void virt_powerdown_req(Notifier *n, void *opaque)
546 {
547     /* use gpio Pin 3 for power button event */
548     qemu_set_irq(qdev_get_gpio_in(pl061_dev, 3), 1);
549 }
550 
551 static Notifier virt_system_powerdown_notifier = {
552     .notify = virt_powerdown_req
553 };
554 
555 static void create_gpio(const VirtBoardInfo *vbi, qemu_irq *pic)
556 {
557     char *nodename;
558     hwaddr base = vbi->memmap[VIRT_GPIO].base;
559     hwaddr size = vbi->memmap[VIRT_GPIO].size;
560     int irq = vbi->irqmap[VIRT_GPIO];
561     const char compat[] = "arm,pl061\0arm,primecell";
562 
563     pl061_dev = sysbus_create_simple("pl061", base, pic[irq]);
564 
565     uint32_t phandle = qemu_fdt_alloc_phandle(vbi->fdt);
566     nodename = g_strdup_printf("/pl061@%" PRIx64, base);
567     qemu_fdt_add_subnode(vbi->fdt, nodename);
568     qemu_fdt_setprop_sized_cells(vbi->fdt, nodename, "reg",
569                                  2, base, 2, size);
570     qemu_fdt_setprop(vbi->fdt, nodename, "compatible", compat, sizeof(compat));
571     qemu_fdt_setprop_cell(vbi->fdt, nodename, "#gpio-cells", 2);
572     qemu_fdt_setprop(vbi->fdt, nodename, "gpio-controller", NULL, 0);
573     qemu_fdt_setprop_cells(vbi->fdt, nodename, "interrupts",
574                            GIC_FDT_IRQ_TYPE_SPI, irq,
575                            GIC_FDT_IRQ_FLAGS_LEVEL_HI);
576     qemu_fdt_setprop_cell(vbi->fdt, nodename, "clocks", vbi->clock_phandle);
577     qemu_fdt_setprop_string(vbi->fdt, nodename, "clock-names", "apb_pclk");
578     qemu_fdt_setprop_cell(vbi->fdt, nodename, "phandle", phandle);
579 
580     qemu_fdt_add_subnode(vbi->fdt, "/gpio-keys");
581     qemu_fdt_setprop_string(vbi->fdt, "/gpio-keys", "compatible", "gpio-keys");
582     qemu_fdt_setprop_cell(vbi->fdt, "/gpio-keys", "#size-cells", 0);
583     qemu_fdt_setprop_cell(vbi->fdt, "/gpio-keys", "#address-cells", 1);
584 
585     qemu_fdt_add_subnode(vbi->fdt, "/gpio-keys/poweroff");
586     qemu_fdt_setprop_string(vbi->fdt, "/gpio-keys/poweroff",
587                             "label", "GPIO Key Poweroff");
588     qemu_fdt_setprop_cell(vbi->fdt, "/gpio-keys/poweroff", "linux,code",
589                           KEY_POWER);
590     qemu_fdt_setprop_cells(vbi->fdt, "/gpio-keys/poweroff",
591                            "gpios", phandle, 3, 0);
592 
593     /* connect powerdown request */
594     qemu_register_powerdown_notifier(&virt_system_powerdown_notifier);
595 
596     g_free(nodename);
597 }
598 
599 static void create_virtio_devices(const VirtBoardInfo *vbi, qemu_irq *pic)
600 {
601     int i;
602     hwaddr size = vbi->memmap[VIRT_MMIO].size;
603 
604     /* We create the transports in forwards order. Since qbus_realize()
605      * prepends (not appends) new child buses, the incrementing loop below will
606      * create a list of virtio-mmio buses with decreasing base addresses.
607      *
608      * When a -device option is processed from the command line,
609      * qbus_find_recursive() picks the next free virtio-mmio bus in forwards
610      * order. The upshot is that -device options in increasing command line
611      * order are mapped to virtio-mmio buses with decreasing base addresses.
612      *
613      * When this code was originally written, that arrangement ensured that the
614      * guest Linux kernel would give the lowest "name" (/dev/vda, eth0, etc) to
615      * the first -device on the command line. (The end-to-end order is a
616      * function of this loop, qbus_realize(), qbus_find_recursive(), and the
617      * guest kernel's name-to-address assignment strategy.)
618      *
619      * Meanwhile, the kernel's traversal seems to have been reversed; see eg.
620      * the message, if not necessarily the code, of commit 70161ff336.
621      * Therefore the loop now establishes the inverse of the original intent.
622      *
623      * Unfortunately, we can't counteract the kernel change by reversing the
624      * loop; it would break existing command lines.
625      *
626      * In any case, the kernel makes no guarantee about the stability of
627      * enumeration order of virtio devices (as demonstrated by it changing
628      * between kernel versions). For reliable and stable identification
629      * of disks users must use UUIDs or similar mechanisms.
630      */
631     for (i = 0; i < NUM_VIRTIO_TRANSPORTS; i++) {
632         int irq = vbi->irqmap[VIRT_MMIO] + i;
633         hwaddr base = vbi->memmap[VIRT_MMIO].base + i * size;
634 
635         sysbus_create_simple("virtio-mmio", base, pic[irq]);
636     }
637 
638     /* We add dtb nodes in reverse order so that they appear in the finished
639      * device tree lowest address first.
640      *
641      * Note that this mapping is independent of the loop above. The previous
642      * loop influences virtio device to virtio transport assignment, whereas
643      * this loop controls how virtio transports are laid out in the dtb.
644      */
645     for (i = NUM_VIRTIO_TRANSPORTS - 1; i >= 0; i--) {
646         char *nodename;
647         int irq = vbi->irqmap[VIRT_MMIO] + i;
648         hwaddr base = vbi->memmap[VIRT_MMIO].base + i * size;
649 
650         nodename = g_strdup_printf("/virtio_mmio@%" PRIx64, base);
651         qemu_fdt_add_subnode(vbi->fdt, nodename);
652         qemu_fdt_setprop_string(vbi->fdt, nodename,
653                                 "compatible", "virtio,mmio");
654         qemu_fdt_setprop_sized_cells(vbi->fdt, nodename, "reg",
655                                      2, base, 2, size);
656         qemu_fdt_setprop_cells(vbi->fdt, nodename, "interrupts",
657                                GIC_FDT_IRQ_TYPE_SPI, irq,
658                                GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
659         g_free(nodename);
660     }
661 }
662 
663 static void create_one_flash(const char *name, hwaddr flashbase,
664                              hwaddr flashsize)
665 {
666     /* Create and map a single flash device. We use the same
667      * parameters as the flash devices on the Versatile Express board.
668      */
669     DriveInfo *dinfo = drive_get_next(IF_PFLASH);
670     DeviceState *dev = qdev_create(NULL, "cfi.pflash01");
671     const uint64_t sectorlength = 256 * 1024;
672 
673     if (dinfo) {
674         qdev_prop_set_drive(dev, "drive", blk_by_legacy_dinfo(dinfo),
675                             &error_abort);
676     }
677 
678     qdev_prop_set_uint32(dev, "num-blocks", flashsize / sectorlength);
679     qdev_prop_set_uint64(dev, "sector-length", sectorlength);
680     qdev_prop_set_uint8(dev, "width", 4);
681     qdev_prop_set_uint8(dev, "device-width", 2);
682     qdev_prop_set_bit(dev, "big-endian", false);
683     qdev_prop_set_uint16(dev, "id0", 0x89);
684     qdev_prop_set_uint16(dev, "id1", 0x18);
685     qdev_prop_set_uint16(dev, "id2", 0x00);
686     qdev_prop_set_uint16(dev, "id3", 0x00);
687     qdev_prop_set_string(dev, "name", name);
688     qdev_init_nofail(dev);
689 
690     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, flashbase);
691 }
692 
693 static void create_flash(const VirtBoardInfo *vbi)
694 {
695     /* Create two flash devices to fill the VIRT_FLASH space in the memmap.
696      * Any file passed via -bios goes in the first of these.
697      */
698     hwaddr flashsize = vbi->memmap[VIRT_FLASH].size / 2;
699     hwaddr flashbase = vbi->memmap[VIRT_FLASH].base;
700     char *nodename;
701 
702     if (bios_name) {
703         char *fn;
704         int image_size;
705 
706         if (drive_get(IF_PFLASH, 0, 0)) {
707             error_report("The contents of the first flash device may be "
708                          "specified with -bios or with -drive if=pflash... "
709                          "but you cannot use both options at once");
710             exit(1);
711         }
712         fn = qemu_find_file(QEMU_FILE_TYPE_BIOS, bios_name);
713         if (!fn) {
714             error_report("Could not find ROM image '%s'", bios_name);
715             exit(1);
716         }
717         image_size = load_image_targphys(fn, flashbase, flashsize);
718         g_free(fn);
719         if (image_size < 0) {
720             error_report("Could not load ROM image '%s'", bios_name);
721             exit(1);
722         }
723     }
724 
725     create_one_flash("virt.flash0", flashbase, flashsize);
726     create_one_flash("virt.flash1", flashbase + flashsize, flashsize);
727 
728     nodename = g_strdup_printf("/flash@%" PRIx64, flashbase);
729     qemu_fdt_add_subnode(vbi->fdt, nodename);
730     qemu_fdt_setprop_string(vbi->fdt, nodename, "compatible", "cfi-flash");
731     qemu_fdt_setprop_sized_cells(vbi->fdt, nodename, "reg",
732                                  2, flashbase, 2, flashsize,
733                                  2, flashbase + flashsize, 2, flashsize);
734     qemu_fdt_setprop_cell(vbi->fdt, nodename, "bank-width", 4);
735     g_free(nodename);
736 }
737 
738 static void create_fw_cfg(const VirtBoardInfo *vbi, AddressSpace *as)
739 {
740     hwaddr base = vbi->memmap[VIRT_FW_CFG].base;
741     hwaddr size = vbi->memmap[VIRT_FW_CFG].size;
742     char *nodename;
743 
744     fw_cfg_init_mem_wide(base + 8, base, 8, base + 16, as);
745 
746     nodename = g_strdup_printf("/fw-cfg@%" PRIx64, base);
747     qemu_fdt_add_subnode(vbi->fdt, nodename);
748     qemu_fdt_setprop_string(vbi->fdt, nodename,
749                             "compatible", "qemu,fw-cfg-mmio");
750     qemu_fdt_setprop_sized_cells(vbi->fdt, nodename, "reg",
751                                  2, base, 2, size);
752     g_free(nodename);
753 }
754 
755 static void create_pcie_irq_map(const VirtBoardInfo *vbi, uint32_t gic_phandle,
756                                 int first_irq, const char *nodename)
757 {
758     int devfn, pin;
759     uint32_t full_irq_map[4 * 4 * 10] = { 0 };
760     uint32_t *irq_map = full_irq_map;
761 
762     for (devfn = 0; devfn <= 0x18; devfn += 0x8) {
763         for (pin = 0; pin < 4; pin++) {
764             int irq_type = GIC_FDT_IRQ_TYPE_SPI;
765             int irq_nr = first_irq + ((pin + PCI_SLOT(devfn)) % PCI_NUM_PINS);
766             int irq_level = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
767             int i;
768 
769             uint32_t map[] = {
770                 devfn << 8, 0, 0,                           /* devfn */
771                 pin + 1,                                    /* PCI pin */
772                 gic_phandle, 0, 0, irq_type, irq_nr, irq_level }; /* GIC irq */
773 
774             /* Convert map to big endian */
775             for (i = 0; i < 10; i++) {
776                 irq_map[i] = cpu_to_be32(map[i]);
777             }
778             irq_map += 10;
779         }
780     }
781 
782     qemu_fdt_setprop(vbi->fdt, nodename, "interrupt-map",
783                      full_irq_map, sizeof(full_irq_map));
784 
785     qemu_fdt_setprop_cells(vbi->fdt, nodename, "interrupt-map-mask",
786                            0x1800, 0, 0, /* devfn (PCI_SLOT(3)) */
787                            0x7           /* PCI irq */);
788 }
789 
790 static void create_pcie(const VirtBoardInfo *vbi, qemu_irq *pic,
791                         bool use_highmem)
792 {
793     hwaddr base_mmio = vbi->memmap[VIRT_PCIE_MMIO].base;
794     hwaddr size_mmio = vbi->memmap[VIRT_PCIE_MMIO].size;
795     hwaddr base_mmio_high = vbi->memmap[VIRT_PCIE_MMIO_HIGH].base;
796     hwaddr size_mmio_high = vbi->memmap[VIRT_PCIE_MMIO_HIGH].size;
797     hwaddr base_pio = vbi->memmap[VIRT_PCIE_PIO].base;
798     hwaddr size_pio = vbi->memmap[VIRT_PCIE_PIO].size;
799     hwaddr base_ecam = vbi->memmap[VIRT_PCIE_ECAM].base;
800     hwaddr size_ecam = vbi->memmap[VIRT_PCIE_ECAM].size;
801     hwaddr base = base_mmio;
802     int nr_pcie_buses = size_ecam / PCIE_MMCFG_SIZE_MIN;
803     int irq = vbi->irqmap[VIRT_PCIE];
804     MemoryRegion *mmio_alias;
805     MemoryRegion *mmio_reg;
806     MemoryRegion *ecam_alias;
807     MemoryRegion *ecam_reg;
808     DeviceState *dev;
809     char *nodename;
810     int i;
811 
812     dev = qdev_create(NULL, TYPE_GPEX_HOST);
813     qdev_init_nofail(dev);
814 
815     /* Map only the first size_ecam bytes of ECAM space */
816     ecam_alias = g_new0(MemoryRegion, 1);
817     ecam_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 0);
818     memory_region_init_alias(ecam_alias, OBJECT(dev), "pcie-ecam",
819                              ecam_reg, 0, size_ecam);
820     memory_region_add_subregion(get_system_memory(), base_ecam, ecam_alias);
821 
822     /* Map the MMIO window into system address space so as to expose
823      * the section of PCI MMIO space which starts at the same base address
824      * (ie 1:1 mapping for that part of PCI MMIO space visible through
825      * the window).
826      */
827     mmio_alias = g_new0(MemoryRegion, 1);
828     mmio_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 1);
829     memory_region_init_alias(mmio_alias, OBJECT(dev), "pcie-mmio",
830                              mmio_reg, base_mmio, size_mmio);
831     memory_region_add_subregion(get_system_memory(), base_mmio, mmio_alias);
832 
833     if (use_highmem) {
834         /* Map high MMIO space */
835         MemoryRegion *high_mmio_alias = g_new0(MemoryRegion, 1);
836 
837         memory_region_init_alias(high_mmio_alias, OBJECT(dev), "pcie-mmio-high",
838                                  mmio_reg, base_mmio_high, size_mmio_high);
839         memory_region_add_subregion(get_system_memory(), base_mmio_high,
840                                     high_mmio_alias);
841     }
842 
843     /* Map IO port space */
844     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 2, base_pio);
845 
846     for (i = 0; i < GPEX_NUM_IRQS; i++) {
847         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i, pic[irq + i]);
848     }
849 
850     nodename = g_strdup_printf("/pcie@%" PRIx64, base);
851     qemu_fdt_add_subnode(vbi->fdt, nodename);
852     qemu_fdt_setprop_string(vbi->fdt, nodename,
853                             "compatible", "pci-host-ecam-generic");
854     qemu_fdt_setprop_string(vbi->fdt, nodename, "device_type", "pci");
855     qemu_fdt_setprop_cell(vbi->fdt, nodename, "#address-cells", 3);
856     qemu_fdt_setprop_cell(vbi->fdt, nodename, "#size-cells", 2);
857     qemu_fdt_setprop_cells(vbi->fdt, nodename, "bus-range", 0,
858                            nr_pcie_buses - 1);
859 
860     if (vbi->v2m_phandle) {
861         qemu_fdt_setprop_cells(vbi->fdt, nodename, "msi-parent",
862                                vbi->v2m_phandle);
863     }
864 
865     qemu_fdt_setprop_sized_cells(vbi->fdt, nodename, "reg",
866                                  2, base_ecam, 2, size_ecam);
867 
868     if (use_highmem) {
869         qemu_fdt_setprop_sized_cells(vbi->fdt, nodename, "ranges",
870                                      1, FDT_PCI_RANGE_IOPORT, 2, 0,
871                                      2, base_pio, 2, size_pio,
872                                      1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
873                                      2, base_mmio, 2, size_mmio,
874                                      1, FDT_PCI_RANGE_MMIO_64BIT,
875                                      2, base_mmio_high,
876                                      2, base_mmio_high, 2, size_mmio_high);
877     } else {
878         qemu_fdt_setprop_sized_cells(vbi->fdt, nodename, "ranges",
879                                      1, FDT_PCI_RANGE_IOPORT, 2, 0,
880                                      2, base_pio, 2, size_pio,
881                                      1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
882                                      2, base_mmio, 2, size_mmio);
883     }
884 
885     qemu_fdt_setprop_cell(vbi->fdt, nodename, "#interrupt-cells", 1);
886     create_pcie_irq_map(vbi, vbi->gic_phandle, irq, nodename);
887 
888     g_free(nodename);
889 }
890 
891 static void create_platform_bus(VirtBoardInfo *vbi, qemu_irq *pic)
892 {
893     DeviceState *dev;
894     SysBusDevice *s;
895     int i;
896     ARMPlatformBusFDTParams *fdt_params = g_new(ARMPlatformBusFDTParams, 1);
897     MemoryRegion *sysmem = get_system_memory();
898 
899     platform_bus_params.platform_bus_base = vbi->memmap[VIRT_PLATFORM_BUS].base;
900     platform_bus_params.platform_bus_size = vbi->memmap[VIRT_PLATFORM_BUS].size;
901     platform_bus_params.platform_bus_first_irq = vbi->irqmap[VIRT_PLATFORM_BUS];
902     platform_bus_params.platform_bus_num_irqs = PLATFORM_BUS_NUM_IRQS;
903 
904     fdt_params->system_params = &platform_bus_params;
905     fdt_params->binfo = &vbi->bootinfo;
906     fdt_params->intc = "/intc";
907     /*
908      * register a machine init done notifier that creates the device tree
909      * nodes of the platform bus and its children dynamic sysbus devices
910      */
911     arm_register_platform_bus_fdt_creator(fdt_params);
912 
913     dev = qdev_create(NULL, TYPE_PLATFORM_BUS_DEVICE);
914     dev->id = TYPE_PLATFORM_BUS_DEVICE;
915     qdev_prop_set_uint32(dev, "num_irqs",
916         platform_bus_params.platform_bus_num_irqs);
917     qdev_prop_set_uint32(dev, "mmio_size",
918         platform_bus_params.platform_bus_size);
919     qdev_init_nofail(dev);
920     s = SYS_BUS_DEVICE(dev);
921 
922     for (i = 0; i < platform_bus_params.platform_bus_num_irqs; i++) {
923         int irqn = platform_bus_params.platform_bus_first_irq + i;
924         sysbus_connect_irq(s, i, pic[irqn]);
925     }
926 
927     memory_region_add_subregion(sysmem,
928                                 platform_bus_params.platform_bus_base,
929                                 sysbus_mmio_get_region(s, 0));
930 }
931 
932 static void *machvirt_dtb(const struct arm_boot_info *binfo, int *fdt_size)
933 {
934     const VirtBoardInfo *board = (const VirtBoardInfo *)binfo;
935 
936     *fdt_size = board->fdt_size;
937     return board->fdt;
938 }
939 
940 static void virt_build_smbios(VirtGuestInfo *guest_info)
941 {
942     FWCfgState *fw_cfg = guest_info->fw_cfg;
943     uint8_t *smbios_tables, *smbios_anchor;
944     size_t smbios_tables_len, smbios_anchor_len;
945     const char *product = "QEMU Virtual Machine";
946 
947     if (!fw_cfg) {
948         return;
949     }
950 
951     if (kvm_enabled()) {
952         product = "KVM Virtual Machine";
953     }
954 
955     smbios_set_defaults("QEMU", product,
956                         "1.0", false, true, SMBIOS_ENTRY_POINT_30);
957 
958     smbios_get_tables(NULL, 0, &smbios_tables, &smbios_tables_len,
959                       &smbios_anchor, &smbios_anchor_len);
960 
961     if (smbios_anchor) {
962         fw_cfg_add_file(fw_cfg, "etc/smbios/smbios-tables",
963                         smbios_tables, smbios_tables_len);
964         fw_cfg_add_file(fw_cfg, "etc/smbios/smbios-anchor",
965                         smbios_anchor, smbios_anchor_len);
966     }
967 }
968 
969 static
970 void virt_guest_info_machine_done(Notifier *notifier, void *data)
971 {
972     VirtGuestInfoState *guest_info_state = container_of(notifier,
973                                               VirtGuestInfoState, machine_done);
974     virt_acpi_setup(&guest_info_state->info);
975     virt_build_smbios(&guest_info_state->info);
976 }
977 
978 static void machvirt_init(MachineState *machine)
979 {
980     VirtMachineState *vms = VIRT_MACHINE(machine);
981     qemu_irq pic[NUM_IRQS];
982     MemoryRegion *sysmem = get_system_memory();
983     int gic_version = vms->gic_version;
984     int n, max_cpus;
985     MemoryRegion *ram = g_new(MemoryRegion, 1);
986     const char *cpu_model = machine->cpu_model;
987     VirtBoardInfo *vbi;
988     VirtGuestInfoState *guest_info_state = g_malloc0(sizeof *guest_info_state);
989     VirtGuestInfo *guest_info = &guest_info_state->info;
990     char **cpustr;
991 
992     if (!cpu_model) {
993         cpu_model = "cortex-a15";
994     }
995 
996     /* We can probe only here because during property set
997      * KVM is not available yet
998      */
999     if (!gic_version) {
1000         gic_version = kvm_arm_vgic_probe();
1001         if (!gic_version) {
1002             error_report("Unable to determine GIC version supported by host");
1003             error_printf("KVM acceleration is probably not supported\n");
1004             exit(1);
1005         }
1006     }
1007 
1008     /* Separate the actual CPU model name from any appended features */
1009     cpustr = g_strsplit(cpu_model, ",", 2);
1010 
1011     vbi = find_machine_info(cpustr[0]);
1012 
1013     if (!vbi) {
1014         error_report("mach-virt: CPU %s not supported", cpustr[0]);
1015         exit(1);
1016     }
1017 
1018     /* The maximum number of CPUs depends on the GIC version, or on how
1019      * many redistributors we can fit into the memory map.
1020      */
1021     if (gic_version == 3) {
1022         max_cpus = vbi->memmap[VIRT_GIC_REDIST].size / 0x20000;
1023     } else {
1024         max_cpus = GIC_NCPU;
1025     }
1026 
1027     if (smp_cpus > max_cpus) {
1028         error_report("Number of SMP CPUs requested (%d) exceeds max CPUs "
1029                      "supported by machine 'mach-virt' (%d)",
1030                      smp_cpus, max_cpus);
1031         exit(1);
1032     }
1033 
1034     vbi->smp_cpus = smp_cpus;
1035 
1036     if (machine->ram_size > vbi->memmap[VIRT_MEM].size) {
1037         error_report("mach-virt: cannot model more than 30GB RAM");
1038         exit(1);
1039     }
1040 
1041     create_fdt(vbi);
1042 
1043     for (n = 0; n < smp_cpus; n++) {
1044         ObjectClass *oc = cpu_class_by_name(TYPE_ARM_CPU, cpustr[0]);
1045         CPUClass *cc = CPU_CLASS(oc);
1046         Object *cpuobj;
1047         Error *err = NULL;
1048         char *cpuopts = g_strdup(cpustr[1]);
1049 
1050         if (!oc) {
1051             error_report("Unable to find CPU definition");
1052             exit(1);
1053         }
1054         cpuobj = object_new(object_class_get_name(oc));
1055 
1056         /* Handle any CPU options specified by the user */
1057         cc->parse_features(CPU(cpuobj), cpuopts, &err);
1058         g_free(cpuopts);
1059         if (err) {
1060             error_report_err(err);
1061             exit(1);
1062         }
1063 
1064         if (!vms->secure) {
1065             object_property_set_bool(cpuobj, false, "has_el3", NULL);
1066         }
1067 
1068         object_property_set_int(cpuobj, QEMU_PSCI_CONDUIT_HVC, "psci-conduit",
1069                                 NULL);
1070 
1071         /* Secondary CPUs start in PSCI powered-down state */
1072         if (n > 0) {
1073             object_property_set_bool(cpuobj, true, "start-powered-off", NULL);
1074         }
1075 
1076         if (object_property_find(cpuobj, "reset-cbar", NULL)) {
1077             object_property_set_int(cpuobj, vbi->memmap[VIRT_CPUPERIPHS].base,
1078                                     "reset-cbar", &error_abort);
1079         }
1080 
1081         object_property_set_bool(cpuobj, true, "realized", NULL);
1082     }
1083     g_strfreev(cpustr);
1084     fdt_add_timer_nodes(vbi, gic_version);
1085     fdt_add_cpu_nodes(vbi);
1086     fdt_add_psci_node(vbi);
1087 
1088     memory_region_allocate_system_memory(ram, NULL, "mach-virt.ram",
1089                                          machine->ram_size);
1090     memory_region_add_subregion(sysmem, vbi->memmap[VIRT_MEM].base, ram);
1091 
1092     create_flash(vbi);
1093 
1094     create_gic(vbi, pic, gic_version, vms->secure);
1095 
1096     create_uart(vbi, pic);
1097 
1098     create_rtc(vbi, pic);
1099 
1100     create_pcie(vbi, pic, vms->highmem);
1101 
1102     create_gpio(vbi, pic);
1103 
1104     /* Create mmio transports, so the user can create virtio backends
1105      * (which will be automatically plugged in to the transports). If
1106      * no backend is created the transport will just sit harmlessly idle.
1107      */
1108     create_virtio_devices(vbi, pic);
1109 
1110     create_fw_cfg(vbi, &address_space_memory);
1111     rom_set_fw(fw_cfg_find());
1112 
1113     guest_info->smp_cpus = smp_cpus;
1114     guest_info->fw_cfg = fw_cfg_find();
1115     guest_info->memmap = vbi->memmap;
1116     guest_info->irqmap = vbi->irqmap;
1117     guest_info->use_highmem = vms->highmem;
1118     guest_info->gic_version = gic_version;
1119     guest_info_state->machine_done.notify = virt_guest_info_machine_done;
1120     qemu_add_machine_init_done_notifier(&guest_info_state->machine_done);
1121 
1122     vbi->bootinfo.ram_size = machine->ram_size;
1123     vbi->bootinfo.kernel_filename = machine->kernel_filename;
1124     vbi->bootinfo.kernel_cmdline = machine->kernel_cmdline;
1125     vbi->bootinfo.initrd_filename = machine->initrd_filename;
1126     vbi->bootinfo.nb_cpus = smp_cpus;
1127     vbi->bootinfo.board_id = -1;
1128     vbi->bootinfo.loader_start = vbi->memmap[VIRT_MEM].base;
1129     vbi->bootinfo.get_dtb = machvirt_dtb;
1130     vbi->bootinfo.firmware_loaded = bios_name || drive_get(IF_PFLASH, 0, 0);
1131     arm_load_kernel(ARM_CPU(first_cpu), &vbi->bootinfo);
1132 
1133     /*
1134      * arm_load_kernel machine init done notifier registration must
1135      * happen before the platform_bus_create call. In this latter,
1136      * another notifier is registered which adds platform bus nodes.
1137      * Notifiers are executed in registration reverse order.
1138      */
1139     create_platform_bus(vbi, pic);
1140 }
1141 
1142 static bool virt_get_secure(Object *obj, Error **errp)
1143 {
1144     VirtMachineState *vms = VIRT_MACHINE(obj);
1145 
1146     return vms->secure;
1147 }
1148 
1149 static void virt_set_secure(Object *obj, bool value, Error **errp)
1150 {
1151     VirtMachineState *vms = VIRT_MACHINE(obj);
1152 
1153     vms->secure = value;
1154 }
1155 
1156 static bool virt_get_highmem(Object *obj, Error **errp)
1157 {
1158     VirtMachineState *vms = VIRT_MACHINE(obj);
1159 
1160     return vms->highmem;
1161 }
1162 
1163 static void virt_set_highmem(Object *obj, bool value, Error **errp)
1164 {
1165     VirtMachineState *vms = VIRT_MACHINE(obj);
1166 
1167     vms->highmem = value;
1168 }
1169 
1170 static char *virt_get_gic_version(Object *obj, Error **errp)
1171 {
1172     VirtMachineState *vms = VIRT_MACHINE(obj);
1173     const char *val = vms->gic_version == 3 ? "3" : "2";
1174 
1175     return g_strdup(val);
1176 }
1177 
1178 static void virt_set_gic_version(Object *obj, const char *value, Error **errp)
1179 {
1180     VirtMachineState *vms = VIRT_MACHINE(obj);
1181 
1182     if (!strcmp(value, "3")) {
1183         vms->gic_version = 3;
1184     } else if (!strcmp(value, "2")) {
1185         vms->gic_version = 2;
1186     } else if (!strcmp(value, "host")) {
1187         vms->gic_version = 0; /* Will probe later */
1188     } else {
1189         error_report("Invalid gic-version option value");
1190         error_printf("Allowed gic-version values are: 3, 2, host\n");
1191         exit(1);
1192     }
1193 }
1194 
1195 static void virt_instance_init(Object *obj)
1196 {
1197     VirtMachineState *vms = VIRT_MACHINE(obj);
1198 
1199     /* EL3 is disabled by default on virt: this makes us consistent
1200      * between KVM and TCG for this board, and it also allows us to
1201      * boot UEFI blobs which assume no TrustZone support.
1202      */
1203     vms->secure = false;
1204     object_property_add_bool(obj, "secure", virt_get_secure,
1205                              virt_set_secure, NULL);
1206     object_property_set_description(obj, "secure",
1207                                     "Set on/off to enable/disable the ARM "
1208                                     "Security Extensions (TrustZone)",
1209                                     NULL);
1210 
1211     /* High memory is enabled by default */
1212     vms->highmem = true;
1213     object_property_add_bool(obj, "highmem", virt_get_highmem,
1214                              virt_set_highmem, NULL);
1215     object_property_set_description(obj, "highmem",
1216                                     "Set on/off to enable/disable using "
1217                                     "physical address space above 32 bits",
1218                                     NULL);
1219     /* Default GIC type is v2 */
1220     vms->gic_version = 2;
1221     object_property_add_str(obj, "gic-version", virt_get_gic_version,
1222                         virt_set_gic_version, NULL);
1223     object_property_set_description(obj, "gic-version",
1224                                     "Set GIC version. "
1225                                     "Valid values are 2, 3 and host", NULL);
1226 }
1227 
1228 static void virt_class_init(ObjectClass *oc, void *data)
1229 {
1230     MachineClass *mc = MACHINE_CLASS(oc);
1231 
1232     mc->desc = "ARM Virtual Machine",
1233     mc->init = machvirt_init;
1234     /* Start max_cpus at the maximum QEMU supports. We'll further restrict
1235      * it later in machvirt_init, where we have more information about the
1236      * configuration of the particular instance.
1237      */
1238     mc->max_cpus = MAX_CPUMASK_BITS;
1239     mc->has_dynamic_sysbus = true;
1240     mc->block_default_type = IF_VIRTIO;
1241     mc->no_cdrom = 1;
1242     mc->pci_allow_0_address = true;
1243 }
1244 
1245 static const TypeInfo machvirt_info = {
1246     .name = TYPE_VIRT_MACHINE,
1247     .parent = TYPE_MACHINE,
1248     .instance_size = sizeof(VirtMachineState),
1249     .instance_init = virt_instance_init,
1250     .class_size = sizeof(VirtMachineClass),
1251     .class_init = virt_class_init,
1252 };
1253 
1254 static void machvirt_machine_init(void)
1255 {
1256     type_register_static(&machvirt_info);
1257 }
1258 
1259 machine_init(machvirt_machine_init);
1260