xref: /openbmc/qemu/hw/arm/virt.c (revision e452053097371880910c744a5d42ae2df058a4a7)
1 /*
2  * ARM mach-virt emulation
3  *
4  * Copyright (c) 2013 Linaro Limited
5  *
6  * This program is free software; you can redistribute it and/or modify it
7  * under the terms and conditions of the GNU General Public License,
8  * version 2 or later, as published by the Free Software Foundation.
9  *
10  * This program is distributed in the hope it will be useful, but WITHOUT
11  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
13  * more details.
14  *
15  * You should have received a copy of the GNU General Public License along with
16  * this program.  If not, see <http://www.gnu.org/licenses/>.
17  *
18  * Emulate a virtual board which works by passing Linux all the information
19  * it needs about what devices are present via the device tree.
20  * There are some restrictions about what we can do here:
21  *  + we can only present devices whose Linux drivers will work based
22  *    purely on the device tree with no platform data at all
23  *  + we want to present a very stripped-down minimalist platform,
24  *    both because this reduces the security attack surface from the guest
25  *    and also because it reduces our exposure to being broken when
26  *    the kernel updates its device tree bindings and requires further
27  *    information in a device binding that we aren't providing.
28  * This is essentially the same approach kvmtool uses.
29  */
30 
31 #include "qemu/osdep.h"
32 #include "qemu/datadir.h"
33 #include "qemu/units.h"
34 #include "qemu/option.h"
35 #include "monitor/qdev.h"
36 #include "hw/sysbus.h"
37 #include "hw/arm/boot.h"
38 #include "hw/arm/primecell.h"
39 #include "hw/arm/virt.h"
40 #include "hw/block/flash.h"
41 #include "hw/vfio/vfio-calxeda-xgmac.h"
42 #include "hw/vfio/vfio-amd-xgbe.h"
43 #include "hw/display/ramfb.h"
44 #include "net/net.h"
45 #include "system/device_tree.h"
46 #include "system/numa.h"
47 #include "system/runstate.h"
48 #include "system/tpm.h"
49 #include "system/tcg.h"
50 #include "system/kvm.h"
51 #include "system/hvf.h"
52 #include "system/qtest.h"
53 #include "hw/loader.h"
54 #include "qapi/error.h"
55 #include "qemu/bitops.h"
56 #include "qemu/cutils.h"
57 #include "qemu/error-report.h"
58 #include "qemu/module.h"
59 #include "hw/pci-host/gpex.h"
60 #include "hw/pci-bridge/pci_expander_bridge.h"
61 #include "hw/virtio/virtio-pci.h"
62 #include "hw/core/sysbus-fdt.h"
63 #include "hw/platform-bus.h"
64 #include "hw/qdev-properties.h"
65 #include "hw/arm/fdt.h"
66 #include "hw/intc/arm_gic.h"
67 #include "hw/intc/arm_gicv3_common.h"
68 #include "hw/intc/arm_gicv3_its_common.h"
69 #include "hw/irq.h"
70 #include "kvm_arm.h"
71 #include "hvf_arm.h"
72 #include "hw/firmware/smbios.h"
73 #include "qapi/visitor.h"
74 #include "qapi/qapi-visit-common.h"
75 #include "qobject/qlist.h"
76 #include "standard-headers/linux/input.h"
77 #include "hw/arm/smmuv3.h"
78 #include "hw/acpi/acpi.h"
79 #include "hw/acpi/pcihp.h"
80 #include "target/arm/cpu-qom.h"
81 #include "target/arm/internals.h"
82 #include "target/arm/multiprocessing.h"
83 #include "target/arm/gtimer.h"
84 #include "hw/mem/pc-dimm.h"
85 #include "hw/mem/nvdimm.h"
86 #include "hw/acpi/generic_event_device.h"
87 #include "hw/uefi/var-service-api.h"
88 #include "hw/virtio/virtio-md-pci.h"
89 #include "hw/virtio/virtio-iommu.h"
90 #include "hw/char/pl011.h"
91 #include "hw/cxl/cxl.h"
92 #include "hw/cxl/cxl_host.h"
93 #include "qemu/guest-random.h"
94 
95 static GlobalProperty arm_virt_compat[] = {
96     { TYPE_VIRTIO_IOMMU_PCI, "aw-bits", "48" },
97 };
98 static const size_t arm_virt_compat_len = G_N_ELEMENTS(arm_virt_compat);
99 
100 /*
101  * This cannot be called from the virt_machine_class_init() because
102  * TYPE_VIRT_MACHINE is abstract and mc->compat_props g_ptr_array_new()
103  * only is called on virt non abstract class init.
104  */
arm_virt_compat_set(MachineClass * mc)105 static void arm_virt_compat_set(MachineClass *mc)
106 {
107     compat_props_add(mc->compat_props, arm_virt_compat,
108                      arm_virt_compat_len);
109 }
110 
111 #define DEFINE_VIRT_MACHINE_IMPL(latest, ...) \
112     static void MACHINE_VER_SYM(class_init, virt, __VA_ARGS__)( \
113         ObjectClass *oc, \
114         const void *data) \
115     { \
116         MachineClass *mc = MACHINE_CLASS(oc); \
117         arm_virt_compat_set(mc); \
118         MACHINE_VER_SYM(options, virt, __VA_ARGS__)(mc); \
119         mc->desc = "QEMU " MACHINE_VER_STR(__VA_ARGS__) " ARM Virtual Machine"; \
120         MACHINE_VER_DEPRECATION(__VA_ARGS__); \
121         if (latest) { \
122             mc->alias = "virt"; \
123         } \
124     } \
125     static const TypeInfo MACHINE_VER_SYM(info, virt, __VA_ARGS__) = \
126     { \
127         .name = MACHINE_VER_TYPE_NAME("virt", __VA_ARGS__), \
128         .parent = TYPE_VIRT_MACHINE, \
129         .class_init = MACHINE_VER_SYM(class_init, virt, __VA_ARGS__), \
130     }; \
131     static void MACHINE_VER_SYM(register, virt, __VA_ARGS__)(void) \
132     { \
133         MACHINE_VER_DELETION(__VA_ARGS__); \
134         type_register_static(&MACHINE_VER_SYM(info, virt, __VA_ARGS__)); \
135     } \
136     type_init(MACHINE_VER_SYM(register, virt, __VA_ARGS__));
137 
138 #define DEFINE_VIRT_MACHINE_AS_LATEST(major, minor) \
139     DEFINE_VIRT_MACHINE_IMPL(true, major, minor)
140 #define DEFINE_VIRT_MACHINE(major, minor) \
141     DEFINE_VIRT_MACHINE_IMPL(false, major, minor)
142 
143 
144 /* Number of external interrupt lines to configure the GIC with */
145 #define NUM_IRQS 256
146 
147 #define PLATFORM_BUS_NUM_IRQS 64
148 
149 /* Legacy RAM limit in GB (< version 4.0) */
150 #define LEGACY_RAMLIMIT_GB 255
151 #define LEGACY_RAMLIMIT_BYTES (LEGACY_RAMLIMIT_GB * GiB)
152 
153 /* Addresses and sizes of our components.
154  * 0..128MB is space for a flash device so we can run bootrom code such as UEFI.
155  * 128MB..256MB is used for miscellaneous device I/O.
156  * 256MB..1GB is reserved for possible future PCI support (ie where the
157  * PCI memory window will go if we add a PCI host controller).
158  * 1GB and up is RAM (which may happily spill over into the
159  * high memory region beyond 4GB).
160  * This represents a compromise between how much RAM can be given to
161  * a 32 bit VM and leaving space for expansion and in particular for PCI.
162  * Note that devices should generally be placed at multiples of 0x10000,
163  * to accommodate guests using 64K pages.
164  */
165 static const MemMapEntry base_memmap[] = {
166     /* Space up to 0x8000000 is reserved for a boot ROM */
167     [VIRT_FLASH] =              {          0, 0x08000000 },
168     [VIRT_CPUPERIPHS] =         { 0x08000000, 0x00020000 },
169     /* GIC distributor and CPU interfaces sit inside the CPU peripheral space */
170     [VIRT_GIC_DIST] =           { 0x08000000, 0x00010000 },
171     [VIRT_GIC_CPU] =            { 0x08010000, 0x00010000 },
172     [VIRT_GIC_V2M] =            { 0x08020000, 0x00001000 },
173     [VIRT_GIC_HYP] =            { 0x08030000, 0x00010000 },
174     [VIRT_GIC_VCPU] =           { 0x08040000, 0x00010000 },
175     /* The space in between here is reserved for GICv3 CPU/vCPU/HYP */
176     [VIRT_GIC_ITS] =            { 0x08080000, 0x00020000 },
177     /* This redistributor space allows up to 2*64kB*123 CPUs */
178     [VIRT_GIC_REDIST] =         { 0x080A0000, 0x00F60000 },
179     [VIRT_UART0] =              { 0x09000000, 0x00001000 },
180     [VIRT_RTC] =                { 0x09010000, 0x00001000 },
181     [VIRT_FW_CFG] =             { 0x09020000, 0x00000018 },
182     [VIRT_GPIO] =               { 0x09030000, 0x00001000 },
183     [VIRT_UART1] =              { 0x09040000, 0x00001000 },
184     [VIRT_SMMU] =               { 0x09050000, 0x00020000 },
185     [VIRT_PCDIMM_ACPI] =        { 0x09070000, MEMORY_HOTPLUG_IO_LEN },
186     [VIRT_ACPI_GED] =           { 0x09080000, ACPI_GED_EVT_SEL_LEN },
187     [VIRT_NVDIMM_ACPI] =        { 0x09090000, NVDIMM_ACPI_IO_LEN},
188     [VIRT_PVTIME] =             { 0x090a0000, 0x00010000 },
189     [VIRT_SECURE_GPIO] =        { 0x090b0000, 0x00001000 },
190     [VIRT_ACPI_PCIHP] =         { 0x090c0000, ACPI_PCIHP_SIZE },
191     [VIRT_MMIO] =               { 0x0a000000, 0x00000200 },
192     /* ...repeating for a total of NUM_VIRTIO_TRANSPORTS, each of that size */
193     [VIRT_PLATFORM_BUS] =       { 0x0c000000, 0x02000000 },
194     [VIRT_SECURE_MEM] =         { 0x0e000000, 0x01000000 },
195     [VIRT_PCIE_MMIO] =          { 0x10000000, 0x2eff0000 },
196     [VIRT_PCIE_PIO] =           { 0x3eff0000, 0x00010000 },
197     [VIRT_PCIE_ECAM] =          { 0x3f000000, 0x01000000 },
198     /* Actual RAM size depends on initial RAM and device memory settings */
199     [VIRT_MEM] =                { GiB, LEGACY_RAMLIMIT_BYTES },
200 };
201 
202 /* Update the docs for highmem-mmio-size when changing this default */
203 #define DEFAULT_HIGH_PCIE_MMIO_SIZE_GB 512
204 #define DEFAULT_HIGH_PCIE_MMIO_SIZE (DEFAULT_HIGH_PCIE_MMIO_SIZE_GB * GiB)
205 
206 /*
207  * Highmem IO Regions: This memory map is floating, located after the RAM.
208  * Each MemMapEntry base (GPA) will be dynamically computed, depending on the
209  * top of the RAM, so that its base get the same alignment as the size,
210  * ie. a 512GiB entry will be aligned on a 512GiB boundary. If there is
211  * less than 256GiB of RAM, the floating area starts at the 256GiB mark.
212  * Note the extended_memmap is sized so that it eventually also includes the
213  * base_memmap entries (VIRT_HIGH_GIC_REDIST2 index is greater than the last
214  * index of base_memmap).
215  *
216  * The memory map for these Highmem IO Regions can be in legacy or compact
217  * layout, depending on 'compact-highmem' property. With legacy layout, the
218  * PA space for one specific region is always reserved, even if the region
219  * has been disabled or doesn't fit into the PA space. However, the PA space
220  * for the region won't be reserved in these circumstances with compact layout.
221  *
222  * Note that the highmem-mmio-size property will update the high PCIE MMIO size
223  * field in this array.
224  */
225 static MemMapEntry extended_memmap[] = {
226     /* Additional 64 MB redist region (can contain up to 512 redistributors) */
227     [VIRT_HIGH_GIC_REDIST2] =   { 0x0, 64 * MiB },
228     [VIRT_CXL_HOST] =           { 0x0, 64 * KiB * 16 }, /* 16 UID */
229     [VIRT_HIGH_PCIE_ECAM] =     { 0x0, 256 * MiB },
230     /* Second PCIe window */
231     [VIRT_HIGH_PCIE_MMIO] =     { 0x0, DEFAULT_HIGH_PCIE_MMIO_SIZE },
232     /* Any CXL Fixed memory windows come here */
233 };
234 
235 static const int a15irqmap[] = {
236     [VIRT_UART0] = 1,
237     [VIRT_RTC] = 2,
238     [VIRT_PCIE] = 3, /* ... to 6 */
239     [VIRT_GPIO] = 7,
240     [VIRT_UART1] = 8,
241     [VIRT_ACPI_GED] = 9,
242     [VIRT_MMIO] = 16, /* ...to 16 + NUM_VIRTIO_TRANSPORTS - 1 */
243     [VIRT_GIC_V2M] = 48, /* ...to 48 + NUM_GICV2M_SPIS - 1 */
244     [VIRT_SMMU] = 74,    /* ...to 74 + NUM_SMMU_IRQS - 1 */
245     [VIRT_PLATFORM_BUS] = 112, /* ...to 112 + PLATFORM_BUS_NUM_IRQS -1 */
246 };
247 
create_randomness(MachineState * ms,const char * node)248 static void create_randomness(MachineState *ms, const char *node)
249 {
250     struct {
251         uint64_t kaslr;
252         uint8_t rng[32];
253     } seed;
254 
255     if (qemu_guest_getrandom(&seed, sizeof(seed), NULL)) {
256         return;
257     }
258     qemu_fdt_setprop_u64(ms->fdt, node, "kaslr-seed", seed.kaslr);
259     qemu_fdt_setprop(ms->fdt, node, "rng-seed", seed.rng, sizeof(seed.rng));
260 }
261 
262 /*
263  * The CPU object always exposes the NS EL2 virt timer IRQ line,
264  * but we don't want to advertise it to the guest in the dtb or ACPI
265  * table unless it's really going to do something.
266  */
ns_el2_virt_timer_present(void)267 static bool ns_el2_virt_timer_present(void)
268 {
269     ARMCPU *cpu = ARM_CPU(qemu_get_cpu(0));
270     CPUARMState *env = &cpu->env;
271 
272     return arm_feature(env, ARM_FEATURE_AARCH64) &&
273         arm_feature(env, ARM_FEATURE_EL2) && cpu_isar_feature(aa64_vh, cpu);
274 }
275 
create_fdt(VirtMachineState * vms)276 static void create_fdt(VirtMachineState *vms)
277 {
278     MachineState *ms = MACHINE(vms);
279     int nb_numa_nodes = ms->numa_state->num_nodes;
280     void *fdt = create_device_tree(&vms->fdt_size);
281 
282     if (!fdt) {
283         error_report("create_device_tree() failed");
284         exit(1);
285     }
286 
287     ms->fdt = fdt;
288 
289     /* Header */
290     qemu_fdt_setprop_string(fdt, "/", "compatible", "linux,dummy-virt");
291     qemu_fdt_setprop_cell(fdt, "/", "#address-cells", 0x2);
292     qemu_fdt_setprop_cell(fdt, "/", "#size-cells", 0x2);
293     qemu_fdt_setprop_string(fdt, "/", "model", "linux,dummy-virt");
294 
295     /*
296      * For QEMU, all DMA is coherent. Advertising this in the root node
297      * has two benefits:
298      *
299      * - It avoids potential bugs where we forget to mark a DMA
300      *   capable device as being dma-coherent
301      * - It avoids spurious warnings from the Linux kernel about
302      *   devices which can't do DMA at all
303      */
304     qemu_fdt_setprop(fdt, "/", "dma-coherent", NULL, 0);
305 
306     /* /chosen must exist for load_dtb to fill in necessary properties later */
307     qemu_fdt_add_subnode(fdt, "/chosen");
308     if (vms->dtb_randomness) {
309         create_randomness(ms, "/chosen");
310     }
311 
312     if (vms->secure) {
313         qemu_fdt_add_subnode(fdt, "/secure-chosen");
314         if (vms->dtb_randomness) {
315             create_randomness(ms, "/secure-chosen");
316         }
317     }
318 
319     qemu_fdt_add_subnode(fdt, "/aliases");
320 
321     /* Clock node, for the benefit of the UART. The kernel device tree
322      * binding documentation claims the PL011 node clock properties are
323      * optional but in practice if you omit them the kernel refuses to
324      * probe for the device.
325      */
326     vms->clock_phandle = qemu_fdt_alloc_phandle(fdt);
327     qemu_fdt_add_subnode(fdt, "/apb-pclk");
328     qemu_fdt_setprop_string(fdt, "/apb-pclk", "compatible", "fixed-clock");
329     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "#clock-cells", 0x0);
330     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "clock-frequency", 24000000);
331     qemu_fdt_setprop_string(fdt, "/apb-pclk", "clock-output-names",
332                                 "clk24mhz");
333     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "phandle", vms->clock_phandle);
334 
335     if (nb_numa_nodes > 0 && ms->numa_state->have_numa_distance) {
336         int size = nb_numa_nodes * nb_numa_nodes * 3 * sizeof(uint32_t);
337         uint32_t *matrix = g_malloc0(size);
338         int idx, i, j;
339 
340         for (i = 0; i < nb_numa_nodes; i++) {
341             for (j = 0; j < nb_numa_nodes; j++) {
342                 idx = (i * nb_numa_nodes + j) * 3;
343                 matrix[idx + 0] = cpu_to_be32(i);
344                 matrix[idx + 1] = cpu_to_be32(j);
345                 matrix[idx + 2] =
346                     cpu_to_be32(ms->numa_state->nodes[i].distance[j]);
347             }
348         }
349 
350         qemu_fdt_add_subnode(fdt, "/distance-map");
351         qemu_fdt_setprop_string(fdt, "/distance-map", "compatible",
352                                 "numa-distance-map-v1");
353         qemu_fdt_setprop(fdt, "/distance-map", "distance-matrix",
354                          matrix, size);
355         g_free(matrix);
356     }
357 }
358 
fdt_add_timer_nodes(const VirtMachineState * vms)359 static void fdt_add_timer_nodes(const VirtMachineState *vms)
360 {
361     /* On real hardware these interrupts are level-triggered.
362      * On KVM they were edge-triggered before host kernel version 4.4,
363      * and level-triggered afterwards.
364      * On emulated QEMU they are level-triggered.
365      *
366      * Getting the DTB info about them wrong is awkward for some
367      * guest kernels:
368      *  pre-4.8 ignore the DT and leave the interrupt configured
369      *   with whatever the GIC reset value (or the bootloader) left it at
370      *  4.8 before rc6 honour the incorrect data by programming it back
371      *   into the GIC, causing problems
372      *  4.8rc6 and later ignore the DT and always write "level triggered"
373      *   into the GIC
374      *
375      * For backwards-compatibility, virt-2.8 and earlier will continue
376      * to say these are edge-triggered, but later machines will report
377      * the correct information.
378      */
379     ARMCPU *armcpu;
380     uint32_t irqflags = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
381     MachineState *ms = MACHINE(vms);
382 
383     if (vms->gic_version == VIRT_GIC_VERSION_2) {
384         irqflags = deposit32(irqflags, GIC_FDT_IRQ_PPI_CPU_START,
385                              GIC_FDT_IRQ_PPI_CPU_WIDTH,
386                              (1 << MACHINE(vms)->smp.cpus) - 1);
387     }
388 
389     qemu_fdt_add_subnode(ms->fdt, "/timer");
390 
391     armcpu = ARM_CPU(qemu_get_cpu(0));
392     if (arm_feature(&armcpu->env, ARM_FEATURE_V8)) {
393         const char compat[] = "arm,armv8-timer\0arm,armv7-timer";
394         qemu_fdt_setprop(ms->fdt, "/timer", "compatible",
395                          compat, sizeof(compat));
396     } else {
397         qemu_fdt_setprop_string(ms->fdt, "/timer", "compatible",
398                                 "arm,armv7-timer");
399     }
400     qemu_fdt_setprop(ms->fdt, "/timer", "always-on", NULL, 0);
401     if (vms->ns_el2_virt_timer_irq) {
402         qemu_fdt_setprop_cells(ms->fdt, "/timer", "interrupts",
403                                GIC_FDT_IRQ_TYPE_PPI,
404                                INTID_TO_PPI(ARCH_TIMER_S_EL1_IRQ), irqflags,
405                                GIC_FDT_IRQ_TYPE_PPI,
406                                INTID_TO_PPI(ARCH_TIMER_NS_EL1_IRQ), irqflags,
407                                GIC_FDT_IRQ_TYPE_PPI,
408                                INTID_TO_PPI(ARCH_TIMER_VIRT_IRQ), irqflags,
409                                GIC_FDT_IRQ_TYPE_PPI,
410                                INTID_TO_PPI(ARCH_TIMER_NS_EL2_IRQ), irqflags,
411                                GIC_FDT_IRQ_TYPE_PPI,
412                                INTID_TO_PPI(ARCH_TIMER_NS_EL2_VIRT_IRQ), irqflags);
413     } else {
414         qemu_fdt_setprop_cells(ms->fdt, "/timer", "interrupts",
415                                GIC_FDT_IRQ_TYPE_PPI,
416                                INTID_TO_PPI(ARCH_TIMER_S_EL1_IRQ), irqflags,
417                                GIC_FDT_IRQ_TYPE_PPI,
418                                INTID_TO_PPI(ARCH_TIMER_NS_EL1_IRQ), irqflags,
419                                GIC_FDT_IRQ_TYPE_PPI,
420                                INTID_TO_PPI(ARCH_TIMER_VIRT_IRQ), irqflags,
421                                GIC_FDT_IRQ_TYPE_PPI,
422                                INTID_TO_PPI(ARCH_TIMER_NS_EL2_IRQ), irqflags);
423     }
424 }
425 
fdt_add_cpu_nodes(const VirtMachineState * vms)426 static void fdt_add_cpu_nodes(const VirtMachineState *vms)
427 {
428     int cpu;
429     int addr_cells = 1;
430     const MachineState *ms = MACHINE(vms);
431     const VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
432     int smp_cpus = ms->smp.cpus;
433 
434     /*
435      * See Linux Documentation/devicetree/bindings/arm/cpus.yaml
436      * On ARM v8 64-bit systems value should be set to 2,
437      * that corresponds to the MPIDR_EL1 register size.
438      * If MPIDR_EL1[63:32] value is equal to 0 on all CPUs
439      * in the system, #address-cells can be set to 1, since
440      * MPIDR_EL1[63:32] bits are not used for CPUs
441      * identification.
442      *
443      * Here we actually don't know whether our system is 32- or 64-bit one.
444      * The simplest way to go is to examine affinity IDs of all our CPUs. If
445      * at least one of them has Aff3 populated, we set #address-cells to 2.
446      */
447     for (cpu = 0; cpu < smp_cpus; cpu++) {
448         ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
449 
450         if (arm_cpu_mp_affinity(armcpu) & ARM_AFF3_MASK) {
451             addr_cells = 2;
452             break;
453         }
454     }
455 
456     qemu_fdt_add_subnode(ms->fdt, "/cpus");
457     qemu_fdt_setprop_cell(ms->fdt, "/cpus", "#address-cells", addr_cells);
458     qemu_fdt_setprop_cell(ms->fdt, "/cpus", "#size-cells", 0x0);
459 
460     for (cpu = smp_cpus - 1; cpu >= 0; cpu--) {
461         char *nodename = g_strdup_printf("/cpus/cpu@%d", cpu);
462         ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
463         CPUState *cs = CPU(armcpu);
464 
465         qemu_fdt_add_subnode(ms->fdt, nodename);
466         qemu_fdt_setprop_string(ms->fdt, nodename, "device_type", "cpu");
467         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
468                                     armcpu->dtb_compatible);
469 
470         if (vms->psci_conduit != QEMU_PSCI_CONDUIT_DISABLED && smp_cpus > 1) {
471             qemu_fdt_setprop_string(ms->fdt, nodename,
472                                         "enable-method", "psci");
473         }
474 
475         if (addr_cells == 2) {
476             qemu_fdt_setprop_u64(ms->fdt, nodename, "reg",
477                                  arm_cpu_mp_affinity(armcpu));
478         } else {
479             qemu_fdt_setprop_cell(ms->fdt, nodename, "reg",
480                                   arm_cpu_mp_affinity(armcpu));
481         }
482 
483         if (ms->possible_cpus->cpus[cs->cpu_index].props.has_node_id) {
484             qemu_fdt_setprop_cell(ms->fdt, nodename, "numa-node-id",
485                 ms->possible_cpus->cpus[cs->cpu_index].props.node_id);
486         }
487 
488         if (!vmc->no_cpu_topology) {
489             qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle",
490                                   qemu_fdt_alloc_phandle(ms->fdt));
491         }
492 
493         g_free(nodename);
494     }
495 
496     if (!vmc->no_cpu_topology) {
497         /*
498          * Add vCPU topology description through fdt node cpu-map.
499          *
500          * See Linux Documentation/devicetree/bindings/cpu/cpu-topology.txt
501          * In a SMP system, the hierarchy of CPUs can be defined through
502          * four entities that are used to describe the layout of CPUs in
503          * the system: socket/cluster/core/thread.
504          *
505          * A socket node represents the boundary of system physical package
506          * and its child nodes must be one or more cluster nodes. A system
507          * can contain several layers of clustering within a single physical
508          * package and cluster nodes can be contained in parent cluster nodes.
509          *
510          * Note: currently we only support one layer of clustering within
511          * each physical package.
512          */
513         qemu_fdt_add_subnode(ms->fdt, "/cpus/cpu-map");
514 
515         for (cpu = smp_cpus - 1; cpu >= 0; cpu--) {
516             char *cpu_path = g_strdup_printf("/cpus/cpu@%d", cpu);
517             char *map_path;
518 
519             if (ms->smp.threads > 1) {
520                 map_path = g_strdup_printf(
521                     "/cpus/cpu-map/socket%d/cluster%d/core%d/thread%d",
522                     cpu / (ms->smp.clusters * ms->smp.cores * ms->smp.threads),
523                     (cpu / (ms->smp.cores * ms->smp.threads)) % ms->smp.clusters,
524                     (cpu / ms->smp.threads) % ms->smp.cores,
525                     cpu % ms->smp.threads);
526             } else {
527                 map_path = g_strdup_printf(
528                     "/cpus/cpu-map/socket%d/cluster%d/core%d",
529                     cpu / (ms->smp.clusters * ms->smp.cores),
530                     (cpu / ms->smp.cores) % ms->smp.clusters,
531                     cpu % ms->smp.cores);
532             }
533             qemu_fdt_add_path(ms->fdt, map_path);
534             qemu_fdt_setprop_phandle(ms->fdt, map_path, "cpu", cpu_path);
535 
536             g_free(map_path);
537             g_free(cpu_path);
538         }
539     }
540 }
541 
fdt_add_its_gic_node(VirtMachineState * vms)542 static void fdt_add_its_gic_node(VirtMachineState *vms)
543 {
544     char *nodename;
545     MachineState *ms = MACHINE(vms);
546 
547     vms->msi_phandle = qemu_fdt_alloc_phandle(ms->fdt);
548     nodename = g_strdup_printf("/intc/its@%" PRIx64,
549                                vms->memmap[VIRT_GIC_ITS].base);
550     qemu_fdt_add_subnode(ms->fdt, nodename);
551     qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
552                             "arm,gic-v3-its");
553     qemu_fdt_setprop(ms->fdt, nodename, "msi-controller", NULL, 0);
554     qemu_fdt_setprop_cell(ms->fdt, nodename, "#msi-cells", 1);
555     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
556                                  2, vms->memmap[VIRT_GIC_ITS].base,
557                                  2, vms->memmap[VIRT_GIC_ITS].size);
558     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", vms->msi_phandle);
559     g_free(nodename);
560 }
561 
fdt_add_v2m_gic_node(VirtMachineState * vms)562 static void fdt_add_v2m_gic_node(VirtMachineState *vms)
563 {
564     MachineState *ms = MACHINE(vms);
565     char *nodename;
566 
567     nodename = g_strdup_printf("/intc/v2m@%" PRIx64,
568                                vms->memmap[VIRT_GIC_V2M].base);
569     vms->msi_phandle = qemu_fdt_alloc_phandle(ms->fdt);
570     qemu_fdt_add_subnode(ms->fdt, nodename);
571     qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
572                             "arm,gic-v2m-frame");
573     qemu_fdt_setprop(ms->fdt, nodename, "msi-controller", NULL, 0);
574     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
575                                  2, vms->memmap[VIRT_GIC_V2M].base,
576                                  2, vms->memmap[VIRT_GIC_V2M].size);
577     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", vms->msi_phandle);
578     g_free(nodename);
579 }
580 
fdt_add_gic_node(VirtMachineState * vms)581 static void fdt_add_gic_node(VirtMachineState *vms)
582 {
583     MachineState *ms = MACHINE(vms);
584     char *nodename;
585 
586     vms->gic_phandle = qemu_fdt_alloc_phandle(ms->fdt);
587     qemu_fdt_setprop_cell(ms->fdt, "/", "interrupt-parent", vms->gic_phandle);
588 
589     nodename = g_strdup_printf("/intc@%" PRIx64,
590                                vms->memmap[VIRT_GIC_DIST].base);
591     qemu_fdt_add_subnode(ms->fdt, nodename);
592     qemu_fdt_setprop_cell(ms->fdt, nodename, "#interrupt-cells", 3);
593     qemu_fdt_setprop(ms->fdt, nodename, "interrupt-controller", NULL, 0);
594     qemu_fdt_setprop_cell(ms->fdt, nodename, "#address-cells", 0x2);
595     qemu_fdt_setprop_cell(ms->fdt, nodename, "#size-cells", 0x2);
596     qemu_fdt_setprop(ms->fdt, nodename, "ranges", NULL, 0);
597     if (vms->gic_version != VIRT_GIC_VERSION_2) {
598         int nb_redist_regions = virt_gicv3_redist_region_count(vms);
599 
600         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
601                                 "arm,gic-v3");
602 
603         qemu_fdt_setprop_cell(ms->fdt, nodename,
604                               "#redistributor-regions", nb_redist_regions);
605 
606         if (nb_redist_regions == 1) {
607             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
608                                          2, vms->memmap[VIRT_GIC_DIST].base,
609                                          2, vms->memmap[VIRT_GIC_DIST].size,
610                                          2, vms->memmap[VIRT_GIC_REDIST].base,
611                                          2, vms->memmap[VIRT_GIC_REDIST].size);
612         } else {
613             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
614                                  2, vms->memmap[VIRT_GIC_DIST].base,
615                                  2, vms->memmap[VIRT_GIC_DIST].size,
616                                  2, vms->memmap[VIRT_GIC_REDIST].base,
617                                  2, vms->memmap[VIRT_GIC_REDIST].size,
618                                  2, vms->memmap[VIRT_HIGH_GIC_REDIST2].base,
619                                  2, vms->memmap[VIRT_HIGH_GIC_REDIST2].size);
620         }
621 
622         if (vms->virt) {
623             qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
624                                    GIC_FDT_IRQ_TYPE_PPI,
625                                    INTID_TO_PPI(ARCH_GIC_MAINT_IRQ),
626                                    GIC_FDT_IRQ_FLAGS_LEVEL_HI);
627         }
628     } else {
629         /* 'cortex-a15-gic' means 'GIC v2' */
630         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
631                                 "arm,cortex-a15-gic");
632         if (!vms->virt) {
633             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
634                                          2, vms->memmap[VIRT_GIC_DIST].base,
635                                          2, vms->memmap[VIRT_GIC_DIST].size,
636                                          2, vms->memmap[VIRT_GIC_CPU].base,
637                                          2, vms->memmap[VIRT_GIC_CPU].size);
638         } else {
639             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
640                                          2, vms->memmap[VIRT_GIC_DIST].base,
641                                          2, vms->memmap[VIRT_GIC_DIST].size,
642                                          2, vms->memmap[VIRT_GIC_CPU].base,
643                                          2, vms->memmap[VIRT_GIC_CPU].size,
644                                          2, vms->memmap[VIRT_GIC_HYP].base,
645                                          2, vms->memmap[VIRT_GIC_HYP].size,
646                                          2, vms->memmap[VIRT_GIC_VCPU].base,
647                                          2, vms->memmap[VIRT_GIC_VCPU].size);
648             qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
649                                    GIC_FDT_IRQ_TYPE_PPI,
650                                    INTID_TO_PPI(ARCH_GIC_MAINT_IRQ),
651                                    GIC_FDT_IRQ_FLAGS_LEVEL_HI);
652         }
653     }
654 
655     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", vms->gic_phandle);
656     g_free(nodename);
657 }
658 
fdt_add_pmu_nodes(const VirtMachineState * vms)659 static void fdt_add_pmu_nodes(const VirtMachineState *vms)
660 {
661     ARMCPU *armcpu = ARM_CPU(first_cpu);
662     uint32_t irqflags = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
663     MachineState *ms = MACHINE(vms);
664 
665     if (!arm_feature(&armcpu->env, ARM_FEATURE_PMU)) {
666         assert(!object_property_get_bool(OBJECT(armcpu), "pmu", NULL));
667         return;
668     }
669 
670     if (vms->gic_version == VIRT_GIC_VERSION_2) {
671         irqflags = deposit32(irqflags, GIC_FDT_IRQ_PPI_CPU_START,
672                              GIC_FDT_IRQ_PPI_CPU_WIDTH,
673                              (1 << MACHINE(vms)->smp.cpus) - 1);
674     }
675 
676     qemu_fdt_add_subnode(ms->fdt, "/pmu");
677     if (arm_feature(&armcpu->env, ARM_FEATURE_V8)) {
678         const char compat[] = "arm,armv8-pmuv3";
679         qemu_fdt_setprop(ms->fdt, "/pmu", "compatible",
680                          compat, sizeof(compat));
681         qemu_fdt_setprop_cells(ms->fdt, "/pmu", "interrupts",
682                                GIC_FDT_IRQ_TYPE_PPI,
683                                INTID_TO_PPI(VIRTUAL_PMU_IRQ), irqflags);
684     }
685 }
686 
create_acpi_ged(VirtMachineState * vms)687 static inline DeviceState *create_acpi_ged(VirtMachineState *vms)
688 {
689     DeviceState *dev;
690     MachineState *ms = MACHINE(vms);
691     SysBusDevice *sbdev;
692     int irq = vms->irqmap[VIRT_ACPI_GED];
693     uint32_t event = ACPI_GED_PWR_DOWN_EVT;
694     bool acpi_pcihp;
695 
696     if (ms->ram_slots) {
697         event |= ACPI_GED_MEM_HOTPLUG_EVT;
698     }
699 
700     if (ms->nvdimms_state->is_enabled) {
701         event |= ACPI_GED_NVDIMM_HOTPLUG_EVT;
702     }
703 
704     dev = qdev_new(TYPE_ACPI_GED);
705     qdev_prop_set_uint32(dev, "ged-event", event);
706     object_property_set_link(OBJECT(dev), "bus", OBJECT(vms->bus), &error_abort);
707     sbdev = SYS_BUS_DEVICE(dev);
708     sysbus_realize_and_unref(sbdev, &error_fatal);
709 
710     sysbus_mmio_map_name(sbdev, TYPE_ACPI_GED, vms->memmap[VIRT_ACPI_GED].base);
711     sysbus_mmio_map_name(sbdev, ACPI_MEMHP_REGION_NAME,
712                          vms->memmap[VIRT_PCDIMM_ACPI].base);
713 
714     acpi_pcihp = object_property_get_bool(OBJECT(dev),
715                                           ACPI_PM_PROP_ACPI_PCIHP_BRIDGE, NULL);
716 
717     if (acpi_pcihp) {
718         int pcihp_region_index;
719 
720         pcihp_region_index = sysbus_mmio_map_name(sbdev, ACPI_PCIHP_REGION_NAME,
721                                                   vms->memmap[VIRT_ACPI_PCIHP].base);
722         assert(pcihp_region_index >= 0);
723     }
724 
725     sysbus_connect_irq(sbdev, 0, qdev_get_gpio_in(vms->gic, irq));
726 
727     return dev;
728 }
729 
create_its(VirtMachineState * vms)730 static void create_its(VirtMachineState *vms)
731 {
732     DeviceState *dev;
733 
734     assert(vms->its);
735     if (!kvm_irqchip_in_kernel() && !vms->tcg_its) {
736         /*
737          * Do nothing if ITS is neither supported by the host nor emulated by
738          * the machine.
739          */
740         return;
741     }
742 
743     dev = qdev_new(its_class_name());
744 
745     object_property_set_link(OBJECT(dev), "parent-gicv3", OBJECT(vms->gic),
746                              &error_abort);
747     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
748     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_ITS].base);
749 
750     fdt_add_its_gic_node(vms);
751     vms->msi_controller = VIRT_MSI_CTRL_ITS;
752 }
753 
create_v2m(VirtMachineState * vms)754 static void create_v2m(VirtMachineState *vms)
755 {
756     int i;
757     int irq = vms->irqmap[VIRT_GIC_V2M];
758     DeviceState *dev;
759 
760     dev = qdev_new("arm-gicv2m");
761     qdev_prop_set_uint32(dev, "base-spi", irq);
762     qdev_prop_set_uint32(dev, "num-spi", NUM_GICV2M_SPIS);
763     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
764     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_V2M].base);
765 
766     for (i = 0; i < NUM_GICV2M_SPIS; i++) {
767         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
768                            qdev_get_gpio_in(vms->gic, irq + i));
769     }
770 
771     fdt_add_v2m_gic_node(vms);
772     vms->msi_controller = VIRT_MSI_CTRL_GICV2M;
773 }
774 
775 /*
776  * If the CPU has FEAT_NMI, then turn on the NMI support in the GICv3 too.
777  * It's permitted to have a configuration with NMI in the CPU (and thus the
778  * GICv3 CPU interface) but not in the distributor/redistributors, but it's
779  * not very useful.
780  */
gicv3_nmi_present(VirtMachineState * vms)781 static bool gicv3_nmi_present(VirtMachineState *vms)
782 {
783     ARMCPU *cpu = ARM_CPU(qemu_get_cpu(0));
784 
785     return tcg_enabled() && cpu_isar_feature(aa64_nmi, cpu) &&
786            (vms->gic_version != VIRT_GIC_VERSION_2);
787 }
788 
create_gic(VirtMachineState * vms,MemoryRegion * mem)789 static void create_gic(VirtMachineState *vms, MemoryRegion *mem)
790 {
791     MachineState *ms = MACHINE(vms);
792     /* We create a standalone GIC */
793     SysBusDevice *gicbusdev;
794     const char *gictype;
795     int i;
796     unsigned int smp_cpus = ms->smp.cpus;
797     uint32_t nb_redist_regions = 0;
798     int revision;
799 
800     if (vms->gic_version == VIRT_GIC_VERSION_2) {
801         gictype = gic_class_name();
802     } else {
803         gictype = gicv3_class_name();
804     }
805 
806     switch (vms->gic_version) {
807     case VIRT_GIC_VERSION_2:
808         revision = 2;
809         break;
810     case VIRT_GIC_VERSION_3:
811         revision = 3;
812         break;
813     case VIRT_GIC_VERSION_4:
814         revision = 4;
815         break;
816     default:
817         g_assert_not_reached();
818     }
819 
820     if (kvm_enabled() && vms->virt &&
821         (revision != 3 || !kvm_irqchip_in_kernel())) {
822         error_report("KVM EL2 is only supported with in-kernel GICv3");
823         exit(1);
824     }
825 
826     vms->gic = qdev_new(gictype);
827     qdev_prop_set_uint32(vms->gic, "revision", revision);
828     qdev_prop_set_uint32(vms->gic, "num-cpu", smp_cpus);
829     /* Note that the num-irq property counts both internal and external
830      * interrupts; there are always 32 of the former (mandated by GIC spec).
831      */
832     qdev_prop_set_uint32(vms->gic, "num-irq", NUM_IRQS + 32);
833     if (!kvm_irqchip_in_kernel()) {
834         qdev_prop_set_bit(vms->gic, "has-security-extensions", vms->secure);
835     }
836 
837     if (vms->gic_version != VIRT_GIC_VERSION_2) {
838         QList *redist_region_count;
839         uint32_t redist0_capacity = virt_redist_capacity(vms, VIRT_GIC_REDIST);
840         uint32_t redist0_count = MIN(smp_cpus, redist0_capacity);
841 
842         nb_redist_regions = virt_gicv3_redist_region_count(vms);
843 
844         redist_region_count = qlist_new();
845         qlist_append_int(redist_region_count, redist0_count);
846         if (nb_redist_regions == 2) {
847             uint32_t redist1_capacity =
848                 virt_redist_capacity(vms, VIRT_HIGH_GIC_REDIST2);
849 
850             qlist_append_int(redist_region_count,
851                 MIN(smp_cpus - redist0_count, redist1_capacity));
852         }
853         qdev_prop_set_array(vms->gic, "redist-region-count",
854                             redist_region_count);
855 
856         if (!kvm_irqchip_in_kernel()) {
857             if (vms->tcg_its) {
858                 object_property_set_link(OBJECT(vms->gic), "sysmem",
859                                          OBJECT(mem), &error_fatal);
860                 qdev_prop_set_bit(vms->gic, "has-lpi", true);
861             }
862         } else if (vms->virt) {
863             qdev_prop_set_uint32(vms->gic, "maintenance-interrupt-id",
864                                  ARCH_GIC_MAINT_IRQ);
865         }
866     } else {
867         if (!kvm_irqchip_in_kernel()) {
868             qdev_prop_set_bit(vms->gic, "has-virtualization-extensions",
869                               vms->virt);
870         }
871     }
872 
873     if (gicv3_nmi_present(vms)) {
874         qdev_prop_set_bit(vms->gic, "has-nmi", true);
875     }
876 
877     gicbusdev = SYS_BUS_DEVICE(vms->gic);
878     sysbus_realize_and_unref(gicbusdev, &error_fatal);
879     sysbus_mmio_map(gicbusdev, 0, vms->memmap[VIRT_GIC_DIST].base);
880     if (vms->gic_version != VIRT_GIC_VERSION_2) {
881         sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_REDIST].base);
882         if (nb_redist_regions == 2) {
883             sysbus_mmio_map(gicbusdev, 2,
884                             vms->memmap[VIRT_HIGH_GIC_REDIST2].base);
885         }
886     } else {
887         sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_CPU].base);
888         if (vms->virt) {
889             sysbus_mmio_map(gicbusdev, 2, vms->memmap[VIRT_GIC_HYP].base);
890             sysbus_mmio_map(gicbusdev, 3, vms->memmap[VIRT_GIC_VCPU].base);
891         }
892     }
893 
894     /* Wire the outputs from each CPU's generic timer and the GICv3
895      * maintenance interrupt signal to the appropriate GIC PPI inputs,
896      * and the GIC's IRQ/FIQ/VIRQ/VFIQ/NMI/VINMI interrupt outputs to the
897      * CPU's inputs.
898      */
899     for (i = 0; i < smp_cpus; i++) {
900         DeviceState *cpudev = DEVICE(qemu_get_cpu(i));
901         int intidbase = NUM_IRQS + i * GIC_INTERNAL;
902         /* Mapping from the output timer irq lines from the CPU to the
903          * GIC PPI inputs we use for the virt board.
904          */
905         const int timer_irq[] = {
906             [GTIMER_PHYS] = ARCH_TIMER_NS_EL1_IRQ,
907             [GTIMER_VIRT] = ARCH_TIMER_VIRT_IRQ,
908             [GTIMER_HYP]  = ARCH_TIMER_NS_EL2_IRQ,
909             [GTIMER_SEC]  = ARCH_TIMER_S_EL1_IRQ,
910             [GTIMER_HYPVIRT] = ARCH_TIMER_NS_EL2_VIRT_IRQ,
911             [GTIMER_S_EL2_PHYS] = ARCH_TIMER_S_EL2_IRQ,
912             [GTIMER_S_EL2_VIRT] = ARCH_TIMER_S_EL2_VIRT_IRQ,
913         };
914 
915         for (unsigned irq = 0; irq < ARRAY_SIZE(timer_irq); irq++) {
916             qdev_connect_gpio_out(cpudev, irq,
917                                   qdev_get_gpio_in(vms->gic,
918                                                    intidbase + timer_irq[irq]));
919         }
920 
921         if (vms->gic_version != VIRT_GIC_VERSION_2) {
922             qemu_irq irq = qdev_get_gpio_in(vms->gic,
923                                             intidbase + ARCH_GIC_MAINT_IRQ);
924             qdev_connect_gpio_out_named(cpudev, "gicv3-maintenance-interrupt",
925                                         0, irq);
926         } else if (vms->virt) {
927             qemu_irq irq = qdev_get_gpio_in(vms->gic,
928                                             intidbase + ARCH_GIC_MAINT_IRQ);
929             sysbus_connect_irq(gicbusdev, i + 4 * smp_cpus, irq);
930         }
931 
932         qdev_connect_gpio_out_named(cpudev, "pmu-interrupt", 0,
933                                     qdev_get_gpio_in(vms->gic, intidbase
934                                                      + VIRTUAL_PMU_IRQ));
935 
936         sysbus_connect_irq(gicbusdev, i, qdev_get_gpio_in(cpudev, ARM_CPU_IRQ));
937         sysbus_connect_irq(gicbusdev, i + smp_cpus,
938                            qdev_get_gpio_in(cpudev, ARM_CPU_FIQ));
939         sysbus_connect_irq(gicbusdev, i + 2 * smp_cpus,
940                            qdev_get_gpio_in(cpudev, ARM_CPU_VIRQ));
941         sysbus_connect_irq(gicbusdev, i + 3 * smp_cpus,
942                            qdev_get_gpio_in(cpudev, ARM_CPU_VFIQ));
943 
944         if (vms->gic_version != VIRT_GIC_VERSION_2) {
945             sysbus_connect_irq(gicbusdev, i + 4 * smp_cpus,
946                                qdev_get_gpio_in(cpudev, ARM_CPU_NMI));
947             sysbus_connect_irq(gicbusdev, i + 5 * smp_cpus,
948                                qdev_get_gpio_in(cpudev, ARM_CPU_VINMI));
949         }
950     }
951 
952     fdt_add_gic_node(vms);
953 
954     if (vms->gic_version != VIRT_GIC_VERSION_2 && vms->its) {
955         create_its(vms);
956     } else if (vms->gic_version == VIRT_GIC_VERSION_2) {
957         create_v2m(vms);
958     }
959 }
960 
create_uart(const VirtMachineState * vms,int uart,MemoryRegion * mem,Chardev * chr,bool secure)961 static void create_uart(const VirtMachineState *vms, int uart,
962                         MemoryRegion *mem, Chardev *chr, bool secure)
963 {
964     char *nodename;
965     hwaddr base = vms->memmap[uart].base;
966     hwaddr size = vms->memmap[uart].size;
967     int irq = vms->irqmap[uart];
968     const char compat[] = "arm,pl011\0arm,primecell";
969     const char clocknames[] = "uartclk\0apb_pclk";
970     DeviceState *dev = qdev_new(TYPE_PL011);
971     SysBusDevice *s = SYS_BUS_DEVICE(dev);
972     MachineState *ms = MACHINE(vms);
973 
974     qdev_prop_set_chr(dev, "chardev", chr);
975     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
976     memory_region_add_subregion(mem, base,
977                                 sysbus_mmio_get_region(s, 0));
978     sysbus_connect_irq(s, 0, qdev_get_gpio_in(vms->gic, irq));
979 
980     nodename = g_strdup_printf("/pl011@%" PRIx64, base);
981     qemu_fdt_add_subnode(ms->fdt, nodename);
982     /* Note that we can't use setprop_string because of the embedded NUL */
983     qemu_fdt_setprop(ms->fdt, nodename, "compatible",
984                          compat, sizeof(compat));
985     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
986                                      2, base, 2, size);
987     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
988                                GIC_FDT_IRQ_TYPE_SPI, irq,
989                                GIC_FDT_IRQ_FLAGS_LEVEL_HI);
990     qemu_fdt_setprop_cells(ms->fdt, nodename, "clocks",
991                                vms->clock_phandle, vms->clock_phandle);
992     qemu_fdt_setprop(ms->fdt, nodename, "clock-names",
993                          clocknames, sizeof(clocknames));
994 
995     if (uart == VIRT_UART0) {
996         qemu_fdt_setprop_string(ms->fdt, "/chosen", "stdout-path", nodename);
997         qemu_fdt_setprop_string(ms->fdt, "/aliases", "serial0", nodename);
998     } else {
999         qemu_fdt_setprop_string(ms->fdt, "/aliases", "serial1", nodename);
1000     }
1001     if (secure) {
1002         /* Mark as not usable by the normal world */
1003         qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
1004         qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
1005 
1006         qemu_fdt_setprop_string(ms->fdt, "/secure-chosen", "stdout-path",
1007                                 nodename);
1008     }
1009 
1010     g_free(nodename);
1011 }
1012 
create_rtc(const VirtMachineState * vms)1013 static void create_rtc(const VirtMachineState *vms)
1014 {
1015     char *nodename;
1016     hwaddr base = vms->memmap[VIRT_RTC].base;
1017     hwaddr size = vms->memmap[VIRT_RTC].size;
1018     int irq = vms->irqmap[VIRT_RTC];
1019     const char compat[] = "arm,pl031\0arm,primecell";
1020     MachineState *ms = MACHINE(vms);
1021 
1022     sysbus_create_simple("pl031", base, qdev_get_gpio_in(vms->gic, irq));
1023 
1024     nodename = g_strdup_printf("/pl031@%" PRIx64, base);
1025     qemu_fdt_add_subnode(ms->fdt, nodename);
1026     qemu_fdt_setprop(ms->fdt, nodename, "compatible", compat, sizeof(compat));
1027     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1028                                  2, base, 2, size);
1029     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
1030                            GIC_FDT_IRQ_TYPE_SPI, irq,
1031                            GIC_FDT_IRQ_FLAGS_LEVEL_HI);
1032     qemu_fdt_setprop_cell(ms->fdt, nodename, "clocks", vms->clock_phandle);
1033     qemu_fdt_setprop_string(ms->fdt, nodename, "clock-names", "apb_pclk");
1034     g_free(nodename);
1035 }
1036 
1037 static DeviceState *gpio_key_dev;
virt_powerdown_req(Notifier * n,void * opaque)1038 static void virt_powerdown_req(Notifier *n, void *opaque)
1039 {
1040     VirtMachineState *s = container_of(n, VirtMachineState, powerdown_notifier);
1041 
1042     if (s->acpi_dev) {
1043         acpi_send_event(s->acpi_dev, ACPI_POWER_DOWN_STATUS);
1044     } else {
1045         /* use gpio Pin for power button event */
1046         qemu_set_irq(qdev_get_gpio_in(gpio_key_dev, 0), 1);
1047     }
1048 }
1049 
create_gpio_keys(char * fdt,DeviceState * pl061_dev,uint32_t phandle)1050 static void create_gpio_keys(char *fdt, DeviceState *pl061_dev,
1051                              uint32_t phandle)
1052 {
1053     gpio_key_dev = sysbus_create_simple("gpio-key", -1,
1054                                         qdev_get_gpio_in(pl061_dev,
1055                                                          GPIO_PIN_POWER_BUTTON));
1056 
1057     qemu_fdt_add_subnode(fdt, "/gpio-keys");
1058     qemu_fdt_setprop_string(fdt, "/gpio-keys", "compatible", "gpio-keys");
1059 
1060     qemu_fdt_add_subnode(fdt, "/gpio-keys/poweroff");
1061     qemu_fdt_setprop_string(fdt, "/gpio-keys/poweroff",
1062                             "label", "GPIO Key Poweroff");
1063     qemu_fdt_setprop_cell(fdt, "/gpio-keys/poweroff", "linux,code",
1064                           KEY_POWER);
1065     qemu_fdt_setprop_cells(fdt, "/gpio-keys/poweroff",
1066                            "gpios", phandle, GPIO_PIN_POWER_BUTTON, 0);
1067 }
1068 
1069 #define SECURE_GPIO_POWEROFF 0
1070 #define SECURE_GPIO_RESET    1
1071 
create_secure_gpio_pwr(char * fdt,DeviceState * pl061_dev,uint32_t phandle)1072 static void create_secure_gpio_pwr(char *fdt, DeviceState *pl061_dev,
1073                                    uint32_t phandle)
1074 {
1075     DeviceState *gpio_pwr_dev;
1076 
1077     /* gpio-pwr */
1078     gpio_pwr_dev = sysbus_create_simple("gpio-pwr", -1, NULL);
1079 
1080     /* connect secure pl061 to gpio-pwr */
1081     qdev_connect_gpio_out(pl061_dev, SECURE_GPIO_RESET,
1082                           qdev_get_gpio_in_named(gpio_pwr_dev, "reset", 0));
1083     qdev_connect_gpio_out(pl061_dev, SECURE_GPIO_POWEROFF,
1084                           qdev_get_gpio_in_named(gpio_pwr_dev, "shutdown", 0));
1085 
1086     qemu_fdt_add_subnode(fdt, "/gpio-poweroff");
1087     qemu_fdt_setprop_string(fdt, "/gpio-poweroff", "compatible",
1088                             "gpio-poweroff");
1089     qemu_fdt_setprop_cells(fdt, "/gpio-poweroff",
1090                            "gpios", phandle, SECURE_GPIO_POWEROFF, 0);
1091     qemu_fdt_setprop_string(fdt, "/gpio-poweroff", "status", "disabled");
1092     qemu_fdt_setprop_string(fdt, "/gpio-poweroff", "secure-status",
1093                             "okay");
1094 
1095     qemu_fdt_add_subnode(fdt, "/gpio-restart");
1096     qemu_fdt_setprop_string(fdt, "/gpio-restart", "compatible",
1097                             "gpio-restart");
1098     qemu_fdt_setprop_cells(fdt, "/gpio-restart",
1099                            "gpios", phandle, SECURE_GPIO_RESET, 0);
1100     qemu_fdt_setprop_string(fdt, "/gpio-restart", "status", "disabled");
1101     qemu_fdt_setprop_string(fdt, "/gpio-restart", "secure-status",
1102                             "okay");
1103 }
1104 
create_gpio_devices(const VirtMachineState * vms,int gpio,MemoryRegion * mem)1105 static void create_gpio_devices(const VirtMachineState *vms, int gpio,
1106                                 MemoryRegion *mem)
1107 {
1108     char *nodename;
1109     DeviceState *pl061_dev;
1110     hwaddr base = vms->memmap[gpio].base;
1111     hwaddr size = vms->memmap[gpio].size;
1112     int irq = vms->irqmap[gpio];
1113     const char compat[] = "arm,pl061\0arm,primecell";
1114     SysBusDevice *s;
1115     MachineState *ms = MACHINE(vms);
1116 
1117     pl061_dev = qdev_new("pl061");
1118     /* Pull lines down to 0 if not driven by the PL061 */
1119     qdev_prop_set_uint32(pl061_dev, "pullups", 0);
1120     qdev_prop_set_uint32(pl061_dev, "pulldowns", 0xff);
1121     s = SYS_BUS_DEVICE(pl061_dev);
1122     sysbus_realize_and_unref(s, &error_fatal);
1123     memory_region_add_subregion(mem, base, sysbus_mmio_get_region(s, 0));
1124     sysbus_connect_irq(s, 0, qdev_get_gpio_in(vms->gic, irq));
1125 
1126     uint32_t phandle = qemu_fdt_alloc_phandle(ms->fdt);
1127     nodename = g_strdup_printf("/pl061@%" PRIx64, base);
1128     qemu_fdt_add_subnode(ms->fdt, nodename);
1129     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1130                                  2, base, 2, size);
1131     qemu_fdt_setprop(ms->fdt, nodename, "compatible", compat, sizeof(compat));
1132     qemu_fdt_setprop_cell(ms->fdt, nodename, "#gpio-cells", 2);
1133     qemu_fdt_setprop(ms->fdt, nodename, "gpio-controller", NULL, 0);
1134     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
1135                            GIC_FDT_IRQ_TYPE_SPI, irq,
1136                            GIC_FDT_IRQ_FLAGS_LEVEL_HI);
1137     qemu_fdt_setprop_cell(ms->fdt, nodename, "clocks", vms->clock_phandle);
1138     qemu_fdt_setprop_string(ms->fdt, nodename, "clock-names", "apb_pclk");
1139     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", phandle);
1140 
1141     if (gpio != VIRT_GPIO) {
1142         /* Mark as not usable by the normal world */
1143         qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
1144         qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
1145     }
1146     g_free(nodename);
1147 
1148     /* Child gpio devices */
1149     if (gpio == VIRT_GPIO) {
1150         create_gpio_keys(ms->fdt, pl061_dev, phandle);
1151     } else {
1152         create_secure_gpio_pwr(ms->fdt, pl061_dev, phandle);
1153     }
1154 }
1155 
create_virtio_devices(const VirtMachineState * vms)1156 static void create_virtio_devices(const VirtMachineState *vms)
1157 {
1158     int i;
1159     hwaddr size = vms->memmap[VIRT_MMIO].size;
1160     MachineState *ms = MACHINE(vms);
1161 
1162     /* We create the transports in forwards order. Since qbus_realize()
1163      * prepends (not appends) new child buses, the incrementing loop below will
1164      * create a list of virtio-mmio buses with decreasing base addresses.
1165      *
1166      * When a -device option is processed from the command line,
1167      * qbus_find_recursive() picks the next free virtio-mmio bus in forwards
1168      * order. The upshot is that -device options in increasing command line
1169      * order are mapped to virtio-mmio buses with decreasing base addresses.
1170      *
1171      * When this code was originally written, that arrangement ensured that the
1172      * guest Linux kernel would give the lowest "name" (/dev/vda, eth0, etc) to
1173      * the first -device on the command line. (The end-to-end order is a
1174      * function of this loop, qbus_realize(), qbus_find_recursive(), and the
1175      * guest kernel's name-to-address assignment strategy.)
1176      *
1177      * Meanwhile, the kernel's traversal seems to have been reversed; see eg.
1178      * the message, if not necessarily the code, of commit 70161ff336.
1179      * Therefore the loop now establishes the inverse of the original intent.
1180      *
1181      * Unfortunately, we can't counteract the kernel change by reversing the
1182      * loop; it would break existing command lines.
1183      *
1184      * In any case, the kernel makes no guarantee about the stability of
1185      * enumeration order of virtio devices (as demonstrated by it changing
1186      * between kernel versions). For reliable and stable identification
1187      * of disks users must use UUIDs or similar mechanisms.
1188      */
1189     for (i = 0; i < NUM_VIRTIO_TRANSPORTS; i++) {
1190         int irq = vms->irqmap[VIRT_MMIO] + i;
1191         hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
1192 
1193         sysbus_create_simple("virtio-mmio", base,
1194                              qdev_get_gpio_in(vms->gic, irq));
1195     }
1196 
1197     /* We add dtb nodes in reverse order so that they appear in the finished
1198      * device tree lowest address first.
1199      *
1200      * Note that this mapping is independent of the loop above. The previous
1201      * loop influences virtio device to virtio transport assignment, whereas
1202      * this loop controls how virtio transports are laid out in the dtb.
1203      */
1204     for (i = NUM_VIRTIO_TRANSPORTS - 1; i >= 0; i--) {
1205         char *nodename;
1206         int irq = vms->irqmap[VIRT_MMIO] + i;
1207         hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
1208 
1209         nodename = g_strdup_printf("/virtio_mmio@%" PRIx64, base);
1210         qemu_fdt_add_subnode(ms->fdt, nodename);
1211         qemu_fdt_setprop_string(ms->fdt, nodename,
1212                                 "compatible", "virtio,mmio");
1213         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1214                                      2, base, 2, size);
1215         qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
1216                                GIC_FDT_IRQ_TYPE_SPI, irq,
1217                                GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
1218         qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1219         g_free(nodename);
1220     }
1221 }
1222 
1223 #define VIRT_FLASH_SECTOR_SIZE (256 * KiB)
1224 
virt_flash_create1(VirtMachineState * vms,const char * name,const char * alias_prop_name)1225 static PFlashCFI01 *virt_flash_create1(VirtMachineState *vms,
1226                                         const char *name,
1227                                         const char *alias_prop_name)
1228 {
1229     /*
1230      * Create a single flash device.  We use the same parameters as
1231      * the flash devices on the Versatile Express board.
1232      */
1233     DeviceState *dev = qdev_new(TYPE_PFLASH_CFI01);
1234 
1235     qdev_prop_set_uint64(dev, "sector-length", VIRT_FLASH_SECTOR_SIZE);
1236     qdev_prop_set_uint8(dev, "width", 4);
1237     qdev_prop_set_uint8(dev, "device-width", 2);
1238     qdev_prop_set_bit(dev, "big-endian", false);
1239     qdev_prop_set_uint16(dev, "id0", 0x89);
1240     qdev_prop_set_uint16(dev, "id1", 0x18);
1241     qdev_prop_set_uint16(dev, "id2", 0x00);
1242     qdev_prop_set_uint16(dev, "id3", 0x00);
1243     qdev_prop_set_string(dev, "name", name);
1244     object_property_add_child(OBJECT(vms), name, OBJECT(dev));
1245     object_property_add_alias(OBJECT(vms), alias_prop_name,
1246                               OBJECT(dev), "drive");
1247     return PFLASH_CFI01(dev);
1248 }
1249 
virt_flash_create(VirtMachineState * vms)1250 static void virt_flash_create(VirtMachineState *vms)
1251 {
1252     vms->flash[0] = virt_flash_create1(vms, "virt.flash0", "pflash0");
1253     vms->flash[1] = virt_flash_create1(vms, "virt.flash1", "pflash1");
1254 }
1255 
virt_flash_map1(PFlashCFI01 * flash,hwaddr base,hwaddr size,MemoryRegion * sysmem)1256 static void virt_flash_map1(PFlashCFI01 *flash,
1257                             hwaddr base, hwaddr size,
1258                             MemoryRegion *sysmem)
1259 {
1260     DeviceState *dev = DEVICE(flash);
1261 
1262     assert(QEMU_IS_ALIGNED(size, VIRT_FLASH_SECTOR_SIZE));
1263     assert(size / VIRT_FLASH_SECTOR_SIZE <= UINT32_MAX);
1264     qdev_prop_set_uint32(dev, "num-blocks", size / VIRT_FLASH_SECTOR_SIZE);
1265     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1266 
1267     memory_region_add_subregion(sysmem, base,
1268                                 sysbus_mmio_get_region(SYS_BUS_DEVICE(dev),
1269                                                        0));
1270 }
1271 
virt_flash_map(VirtMachineState * vms,MemoryRegion * sysmem,MemoryRegion * secure_sysmem)1272 static void virt_flash_map(VirtMachineState *vms,
1273                            MemoryRegion *sysmem,
1274                            MemoryRegion *secure_sysmem)
1275 {
1276     /*
1277      * Map two flash devices to fill the VIRT_FLASH space in the memmap.
1278      * sysmem is the system memory space. secure_sysmem is the secure view
1279      * of the system, and the first flash device should be made visible only
1280      * there. The second flash device is visible to both secure and nonsecure.
1281      * If sysmem == secure_sysmem this means there is no separate Secure
1282      * address space and both flash devices are generally visible.
1283      */
1284     hwaddr flashsize = vms->memmap[VIRT_FLASH].size / 2;
1285     hwaddr flashbase = vms->memmap[VIRT_FLASH].base;
1286 
1287     virt_flash_map1(vms->flash[0], flashbase, flashsize,
1288                     secure_sysmem);
1289     virt_flash_map1(vms->flash[1], flashbase + flashsize, flashsize,
1290                     sysmem);
1291 }
1292 
virt_flash_fdt(VirtMachineState * vms,MemoryRegion * sysmem,MemoryRegion * secure_sysmem)1293 static void virt_flash_fdt(VirtMachineState *vms,
1294                            MemoryRegion *sysmem,
1295                            MemoryRegion *secure_sysmem)
1296 {
1297     hwaddr flashsize = vms->memmap[VIRT_FLASH].size / 2;
1298     hwaddr flashbase = vms->memmap[VIRT_FLASH].base;
1299     MachineState *ms = MACHINE(vms);
1300     char *nodename;
1301 
1302     if (sysmem == secure_sysmem) {
1303         /* Report both flash devices as a single node in the DT */
1304         nodename = g_strdup_printf("/flash@%" PRIx64, flashbase);
1305         qemu_fdt_add_subnode(ms->fdt, nodename);
1306         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible", "cfi-flash");
1307         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1308                                      2, flashbase, 2, flashsize,
1309                                      2, flashbase + flashsize, 2, flashsize);
1310         qemu_fdt_setprop_cell(ms->fdt, nodename, "bank-width", 4);
1311         g_free(nodename);
1312     } else {
1313         /*
1314          * Report the devices as separate nodes so we can mark one as
1315          * only visible to the secure world.
1316          */
1317         nodename = g_strdup_printf("/secflash@%" PRIx64, flashbase);
1318         qemu_fdt_add_subnode(ms->fdt, nodename);
1319         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible", "cfi-flash");
1320         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1321                                      2, flashbase, 2, flashsize);
1322         qemu_fdt_setprop_cell(ms->fdt, nodename, "bank-width", 4);
1323         qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
1324         qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
1325         g_free(nodename);
1326 
1327         nodename = g_strdup_printf("/flash@%" PRIx64, flashbase + flashsize);
1328         qemu_fdt_add_subnode(ms->fdt, nodename);
1329         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible", "cfi-flash");
1330         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1331                                      2, flashbase + flashsize, 2, flashsize);
1332         qemu_fdt_setprop_cell(ms->fdt, nodename, "bank-width", 4);
1333         g_free(nodename);
1334     }
1335 }
1336 
virt_firmware_init(VirtMachineState * vms,MemoryRegion * sysmem,MemoryRegion * secure_sysmem)1337 static bool virt_firmware_init(VirtMachineState *vms,
1338                                MemoryRegion *sysmem,
1339                                MemoryRegion *secure_sysmem)
1340 {
1341     int i;
1342     const char *bios_name;
1343     BlockBackend *pflash_blk0;
1344 
1345     /* Map legacy -drive if=pflash to machine properties */
1346     for (i = 0; i < ARRAY_SIZE(vms->flash); i++) {
1347         pflash_cfi01_legacy_drive(vms->flash[i],
1348                                   drive_get(IF_PFLASH, 0, i));
1349     }
1350 
1351     virt_flash_map(vms, sysmem, secure_sysmem);
1352 
1353     pflash_blk0 = pflash_cfi01_get_blk(vms->flash[0]);
1354 
1355     bios_name = MACHINE(vms)->firmware;
1356     if (bios_name) {
1357         char *fname;
1358         MemoryRegion *mr;
1359         int image_size;
1360 
1361         if (pflash_blk0) {
1362             error_report("The contents of the first flash device may be "
1363                          "specified with -bios or with -drive if=pflash... "
1364                          "but you cannot use both options at once");
1365             exit(1);
1366         }
1367 
1368         /* Fall back to -bios */
1369 
1370         fname = qemu_find_file(QEMU_FILE_TYPE_BIOS, bios_name);
1371         if (!fname) {
1372             error_report("Could not find ROM image '%s'", bios_name);
1373             exit(1);
1374         }
1375         mr = sysbus_mmio_get_region(SYS_BUS_DEVICE(vms->flash[0]), 0);
1376         image_size = load_image_mr(fname, mr);
1377         g_free(fname);
1378         if (image_size < 0) {
1379             error_report("Could not load ROM image '%s'", bios_name);
1380             exit(1);
1381         }
1382     }
1383 
1384     return pflash_blk0 || bios_name;
1385 }
1386 
create_fw_cfg(const VirtMachineState * vms,AddressSpace * as)1387 static FWCfgState *create_fw_cfg(const VirtMachineState *vms, AddressSpace *as)
1388 {
1389     MachineState *ms = MACHINE(vms);
1390     hwaddr base = vms->memmap[VIRT_FW_CFG].base;
1391     hwaddr size = vms->memmap[VIRT_FW_CFG].size;
1392     FWCfgState *fw_cfg;
1393     char *nodename;
1394 
1395     fw_cfg = fw_cfg_init_mem_wide(base + 8, base, 8, base + 16, as);
1396     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, (uint16_t)ms->smp.cpus);
1397 
1398     nodename = g_strdup_printf("/fw-cfg@%" PRIx64, base);
1399     qemu_fdt_add_subnode(ms->fdt, nodename);
1400     qemu_fdt_setprop_string(ms->fdt, nodename,
1401                             "compatible", "qemu,fw-cfg-mmio");
1402     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1403                                  2, base, 2, size);
1404     qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1405     g_free(nodename);
1406     return fw_cfg;
1407 }
1408 
create_pcie_irq_map(const MachineState * ms,uint32_t gic_phandle,int first_irq,const char * nodename)1409 static void create_pcie_irq_map(const MachineState *ms,
1410                                 uint32_t gic_phandle,
1411                                 int first_irq, const char *nodename)
1412 {
1413     int devfn, pin;
1414     uint32_t full_irq_map[4 * 4 * 10] = { 0 };
1415     uint32_t *irq_map = full_irq_map;
1416 
1417     for (devfn = 0; devfn <= 0x18; devfn += 0x8) {
1418         for (pin = 0; pin < 4; pin++) {
1419             int irq_type = GIC_FDT_IRQ_TYPE_SPI;
1420             int irq_nr = first_irq + ((pin + PCI_SLOT(devfn)) % PCI_NUM_PINS);
1421             int irq_level = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
1422             int i;
1423 
1424             uint32_t map[] = {
1425                 devfn << 8, 0, 0,                           /* devfn */
1426                 pin + 1,                                    /* PCI pin */
1427                 gic_phandle, 0, 0, irq_type, irq_nr, irq_level }; /* GIC irq */
1428 
1429             /* Convert map to big endian */
1430             for (i = 0; i < 10; i++) {
1431                 irq_map[i] = cpu_to_be32(map[i]);
1432             }
1433             irq_map += 10;
1434         }
1435     }
1436 
1437     qemu_fdt_setprop(ms->fdt, nodename, "interrupt-map",
1438                      full_irq_map, sizeof(full_irq_map));
1439 
1440     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupt-map-mask",
1441                            cpu_to_be16(PCI_DEVFN(3, 0)), /* Slot 3 */
1442                            0, 0,
1443                            0x7           /* PCI irq */);
1444 }
1445 
create_smmu(const VirtMachineState * vms,PCIBus * bus)1446 static void create_smmu(const VirtMachineState *vms,
1447                         PCIBus *bus)
1448 {
1449     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1450     char *node;
1451     const char compat[] = "arm,smmu-v3";
1452     int irq =  vms->irqmap[VIRT_SMMU];
1453     int i;
1454     hwaddr base = vms->memmap[VIRT_SMMU].base;
1455     hwaddr size = vms->memmap[VIRT_SMMU].size;
1456     const char irq_names[] = "eventq\0priq\0cmdq-sync\0gerror";
1457     DeviceState *dev;
1458     MachineState *ms = MACHINE(vms);
1459 
1460     if (vms->iommu != VIRT_IOMMU_SMMUV3 || !vms->iommu_phandle) {
1461         return;
1462     }
1463 
1464     dev = qdev_new(TYPE_ARM_SMMUV3);
1465 
1466     if (!vmc->no_nested_smmu) {
1467         object_property_set_str(OBJECT(dev), "stage", "nested", &error_fatal);
1468     }
1469     object_property_set_link(OBJECT(dev), "primary-bus", OBJECT(bus),
1470                              &error_abort);
1471     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1472     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, base);
1473     for (i = 0; i < NUM_SMMU_IRQS; i++) {
1474         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
1475                            qdev_get_gpio_in(vms->gic, irq + i));
1476     }
1477 
1478     node = g_strdup_printf("/smmuv3@%" PRIx64, base);
1479     qemu_fdt_add_subnode(ms->fdt, node);
1480     qemu_fdt_setprop(ms->fdt, node, "compatible", compat, sizeof(compat));
1481     qemu_fdt_setprop_sized_cells(ms->fdt, node, "reg", 2, base, 2, size);
1482 
1483     qemu_fdt_setprop_cells(ms->fdt, node, "interrupts",
1484             GIC_FDT_IRQ_TYPE_SPI, irq    , GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1485             GIC_FDT_IRQ_TYPE_SPI, irq + 1, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1486             GIC_FDT_IRQ_TYPE_SPI, irq + 2, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1487             GIC_FDT_IRQ_TYPE_SPI, irq + 3, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
1488 
1489     qemu_fdt_setprop(ms->fdt, node, "interrupt-names", irq_names,
1490                      sizeof(irq_names));
1491 
1492     qemu_fdt_setprop(ms->fdt, node, "dma-coherent", NULL, 0);
1493 
1494     qemu_fdt_setprop_cell(ms->fdt, node, "#iommu-cells", 1);
1495 
1496     qemu_fdt_setprop_cell(ms->fdt, node, "phandle", vms->iommu_phandle);
1497     g_free(node);
1498 }
1499 
create_virtio_iommu_dt_bindings(VirtMachineState * vms)1500 static void create_virtio_iommu_dt_bindings(VirtMachineState *vms)
1501 {
1502     const char compat[] = "virtio,pci-iommu\0pci1af4,1057";
1503     uint16_t bdf = vms->virtio_iommu_bdf;
1504     MachineState *ms = MACHINE(vms);
1505     char *node;
1506 
1507     vms->iommu_phandle = qemu_fdt_alloc_phandle(ms->fdt);
1508 
1509     node = g_strdup_printf("%s/virtio_iommu@%x,%x", vms->pciehb_nodename,
1510                            PCI_SLOT(bdf), PCI_FUNC(bdf));
1511     qemu_fdt_add_subnode(ms->fdt, node);
1512     qemu_fdt_setprop(ms->fdt, node, "compatible", compat, sizeof(compat));
1513     qemu_fdt_setprop_sized_cells(ms->fdt, node, "reg",
1514                                  1, bdf << 8, 1, 0, 1, 0,
1515                                  1, 0, 1, 0);
1516 
1517     qemu_fdt_setprop_cell(ms->fdt, node, "#iommu-cells", 1);
1518     qemu_fdt_setprop_cell(ms->fdt, node, "phandle", vms->iommu_phandle);
1519     g_free(node);
1520 
1521     if (!vms->default_bus_bypass_iommu) {
1522         qemu_fdt_setprop_cells(ms->fdt, vms->pciehb_nodename, "iommu-map",
1523                                0x0, vms->iommu_phandle, 0x0, bdf,
1524                                bdf + 1, vms->iommu_phandle, bdf + 1,
1525                                0xffff - bdf);
1526     }
1527 }
1528 
create_pcie(VirtMachineState * vms)1529 static void create_pcie(VirtMachineState *vms)
1530 {
1531     hwaddr base_mmio = vms->memmap[VIRT_PCIE_MMIO].base;
1532     hwaddr size_mmio = vms->memmap[VIRT_PCIE_MMIO].size;
1533     hwaddr base_mmio_high = vms->memmap[VIRT_HIGH_PCIE_MMIO].base;
1534     hwaddr size_mmio_high = vms->memmap[VIRT_HIGH_PCIE_MMIO].size;
1535     hwaddr base_pio = vms->memmap[VIRT_PCIE_PIO].base;
1536     hwaddr size_pio = vms->memmap[VIRT_PCIE_PIO].size;
1537     hwaddr base_ecam, size_ecam;
1538     hwaddr base = base_mmio;
1539     int nr_pcie_buses;
1540     int irq = vms->irqmap[VIRT_PCIE];
1541     MemoryRegion *mmio_alias;
1542     MemoryRegion *mmio_reg;
1543     MemoryRegion *ecam_alias;
1544     MemoryRegion *ecam_reg;
1545     DeviceState *dev;
1546     char *nodename;
1547     int i, ecam_id;
1548     PCIHostState *pci;
1549     MachineState *ms = MACHINE(vms);
1550     MachineClass *mc = MACHINE_GET_CLASS(ms);
1551 
1552     dev = qdev_new(TYPE_GPEX_HOST);
1553     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1554 
1555     ecam_id = VIRT_ECAM_ID(vms->highmem_ecam);
1556     base_ecam = vms->memmap[ecam_id].base;
1557     size_ecam = vms->memmap[ecam_id].size;
1558     nr_pcie_buses = size_ecam / PCIE_MMCFG_SIZE_MIN;
1559     /* Map only the first size_ecam bytes of ECAM space */
1560     ecam_alias = g_new0(MemoryRegion, 1);
1561     ecam_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 0);
1562     memory_region_init_alias(ecam_alias, OBJECT(dev), "pcie-ecam",
1563                              ecam_reg, 0, size_ecam);
1564     memory_region_add_subregion(get_system_memory(), base_ecam, ecam_alias);
1565 
1566     /* Map the MMIO window into system address space so as to expose
1567      * the section of PCI MMIO space which starts at the same base address
1568      * (ie 1:1 mapping for that part of PCI MMIO space visible through
1569      * the window).
1570      */
1571     mmio_alias = g_new0(MemoryRegion, 1);
1572     mmio_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 1);
1573     memory_region_init_alias(mmio_alias, OBJECT(dev), "pcie-mmio",
1574                              mmio_reg, base_mmio, size_mmio);
1575     memory_region_add_subregion(get_system_memory(), base_mmio, mmio_alias);
1576 
1577     if (vms->highmem_mmio) {
1578         /* Map high MMIO space */
1579         MemoryRegion *high_mmio_alias = g_new0(MemoryRegion, 1);
1580 
1581         memory_region_init_alias(high_mmio_alias, OBJECT(dev), "pcie-mmio-high",
1582                                  mmio_reg, base_mmio_high, size_mmio_high);
1583         memory_region_add_subregion(get_system_memory(), base_mmio_high,
1584                                     high_mmio_alias);
1585     }
1586 
1587     /* Map IO port space */
1588     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 2, base_pio);
1589 
1590     for (i = 0; i < PCI_NUM_PINS; i++) {
1591         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
1592                            qdev_get_gpio_in(vms->gic, irq + i));
1593         gpex_set_irq_num(GPEX_HOST(dev), i, irq + i);
1594     }
1595 
1596     pci = PCI_HOST_BRIDGE(dev);
1597     pci->bypass_iommu = vms->default_bus_bypass_iommu;
1598     vms->bus = pci->bus;
1599     if (vms->bus) {
1600         pci_init_nic_devices(pci->bus, mc->default_nic);
1601     }
1602 
1603     nodename = vms->pciehb_nodename = g_strdup_printf("/pcie@%" PRIx64, base);
1604     qemu_fdt_add_subnode(ms->fdt, nodename);
1605     qemu_fdt_setprop_string(ms->fdt, nodename,
1606                             "compatible", "pci-host-ecam-generic");
1607     qemu_fdt_setprop_string(ms->fdt, nodename, "device_type", "pci");
1608     qemu_fdt_setprop_cell(ms->fdt, nodename, "#address-cells", 3);
1609     qemu_fdt_setprop_cell(ms->fdt, nodename, "#size-cells", 2);
1610     qemu_fdt_setprop_cell(ms->fdt, nodename, "linux,pci-domain", 0);
1611     qemu_fdt_setprop_cells(ms->fdt, nodename, "bus-range", 0,
1612                            nr_pcie_buses - 1);
1613     qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1614 
1615     if (vms->msi_phandle) {
1616         qemu_fdt_setprop_cells(ms->fdt, nodename, "msi-map",
1617                                0, vms->msi_phandle, 0, 0x10000);
1618     }
1619 
1620     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1621                                  2, base_ecam, 2, size_ecam);
1622 
1623     if (vms->highmem_mmio) {
1624         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "ranges",
1625                                      1, FDT_PCI_RANGE_IOPORT, 2, 0,
1626                                      2, base_pio, 2, size_pio,
1627                                      1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1628                                      2, base_mmio, 2, size_mmio,
1629                                      1, FDT_PCI_RANGE_MMIO_64BIT,
1630                                      2, base_mmio_high,
1631                                      2, base_mmio_high, 2, size_mmio_high);
1632     } else {
1633         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "ranges",
1634                                      1, FDT_PCI_RANGE_IOPORT, 2, 0,
1635                                      2, base_pio, 2, size_pio,
1636                                      1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1637                                      2, base_mmio, 2, size_mmio);
1638     }
1639 
1640     qemu_fdt_setprop_cell(ms->fdt, nodename, "#interrupt-cells", 1);
1641     create_pcie_irq_map(ms, vms->gic_phandle, irq, nodename);
1642 
1643     if (vms->iommu) {
1644         vms->iommu_phandle = qemu_fdt_alloc_phandle(ms->fdt);
1645 
1646         switch (vms->iommu) {
1647         case VIRT_IOMMU_SMMUV3:
1648             create_smmu(vms, vms->bus);
1649             if (!vms->default_bus_bypass_iommu) {
1650                 qemu_fdt_setprop_cells(ms->fdt, nodename, "iommu-map",
1651                                        0x0, vms->iommu_phandle, 0x0, 0x10000);
1652             }
1653             break;
1654         default:
1655             g_assert_not_reached();
1656         }
1657     }
1658 }
1659 
create_cxl_host_reg_region(VirtMachineState * vms)1660 static void create_cxl_host_reg_region(VirtMachineState *vms)
1661 {
1662     MemoryRegion *sysmem = get_system_memory();
1663     MemoryRegion *mr = &vms->cxl_devices_state.host_mr;
1664 
1665     memory_region_init(mr, OBJECT(vms), "cxl_host_reg",
1666                        vms->memmap[VIRT_CXL_HOST].size);
1667     memory_region_add_subregion(sysmem, vms->memmap[VIRT_CXL_HOST].base, mr);
1668     vms->highmem_cxl = true;
1669 }
1670 
create_platform_bus(VirtMachineState * vms)1671 static void create_platform_bus(VirtMachineState *vms)
1672 {
1673     DeviceState *dev;
1674     SysBusDevice *s;
1675     int i;
1676     MemoryRegion *sysmem = get_system_memory();
1677 
1678     dev = qdev_new(TYPE_PLATFORM_BUS_DEVICE);
1679     dev->id = g_strdup(TYPE_PLATFORM_BUS_DEVICE);
1680     qdev_prop_set_uint32(dev, "num_irqs", PLATFORM_BUS_NUM_IRQS);
1681     qdev_prop_set_uint32(dev, "mmio_size", vms->memmap[VIRT_PLATFORM_BUS].size);
1682     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1683     vms->platform_bus_dev = dev;
1684 
1685     s = SYS_BUS_DEVICE(dev);
1686     for (i = 0; i < PLATFORM_BUS_NUM_IRQS; i++) {
1687         int irq = vms->irqmap[VIRT_PLATFORM_BUS] + i;
1688         sysbus_connect_irq(s, i, qdev_get_gpio_in(vms->gic, irq));
1689     }
1690 
1691     memory_region_add_subregion(sysmem,
1692                                 vms->memmap[VIRT_PLATFORM_BUS].base,
1693                                 sysbus_mmio_get_region(s, 0));
1694 }
1695 
create_tag_ram(MemoryRegion * tag_sysmem,hwaddr base,hwaddr size,const char * name)1696 static void create_tag_ram(MemoryRegion *tag_sysmem,
1697                            hwaddr base, hwaddr size,
1698                            const char *name)
1699 {
1700     MemoryRegion *tagram = g_new(MemoryRegion, 1);
1701 
1702     memory_region_init_ram(tagram, NULL, name, size / 32, &error_fatal);
1703     memory_region_add_subregion(tag_sysmem, base / 32, tagram);
1704 }
1705 
create_secure_ram(VirtMachineState * vms,MemoryRegion * secure_sysmem,MemoryRegion * secure_tag_sysmem)1706 static void create_secure_ram(VirtMachineState *vms,
1707                               MemoryRegion *secure_sysmem,
1708                               MemoryRegion *secure_tag_sysmem)
1709 {
1710     MemoryRegion *secram = g_new(MemoryRegion, 1);
1711     char *nodename;
1712     hwaddr base = vms->memmap[VIRT_SECURE_MEM].base;
1713     hwaddr size = vms->memmap[VIRT_SECURE_MEM].size;
1714     MachineState *ms = MACHINE(vms);
1715 
1716     memory_region_init_ram(secram, NULL, "virt.secure-ram", size,
1717                            &error_fatal);
1718     memory_region_add_subregion(secure_sysmem, base, secram);
1719 
1720     nodename = g_strdup_printf("/secram@%" PRIx64, base);
1721     qemu_fdt_add_subnode(ms->fdt, nodename);
1722     qemu_fdt_setprop_string(ms->fdt, nodename, "device_type", "memory");
1723     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg", 2, base, 2, size);
1724     qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
1725     qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
1726 
1727     if (secure_tag_sysmem) {
1728         create_tag_ram(secure_tag_sysmem, base, size, "mach-virt.secure-tag");
1729     }
1730 
1731     g_free(nodename);
1732 }
1733 
machvirt_dtb(const struct arm_boot_info * binfo,int * fdt_size)1734 static void *machvirt_dtb(const struct arm_boot_info *binfo, int *fdt_size)
1735 {
1736     const VirtMachineState *board = container_of(binfo, VirtMachineState,
1737                                                  bootinfo);
1738     MachineState *ms = MACHINE(board);
1739 
1740 
1741     *fdt_size = board->fdt_size;
1742     return ms->fdt;
1743 }
1744 
virt_build_smbios(VirtMachineState * vms)1745 static void virt_build_smbios(VirtMachineState *vms)
1746 {
1747     MachineClass *mc = MACHINE_GET_CLASS(vms);
1748     MachineState *ms = MACHINE(vms);
1749     uint8_t *smbios_tables, *smbios_anchor;
1750     size_t smbios_tables_len, smbios_anchor_len;
1751     struct smbios_phys_mem_area mem_array;
1752     const char *product = "QEMU Virtual Machine";
1753 
1754     if (kvm_enabled()) {
1755         product = "KVM Virtual Machine";
1756     }
1757 
1758     smbios_set_defaults("QEMU", product, mc->name);
1759 
1760     /* build the array of physical mem area from base_memmap */
1761     mem_array.address = vms->memmap[VIRT_MEM].base;
1762     mem_array.length = ms->ram_size;
1763 
1764     smbios_get_tables(ms, SMBIOS_ENTRY_POINT_TYPE_64, &mem_array, 1,
1765                       &smbios_tables, &smbios_tables_len,
1766                       &smbios_anchor, &smbios_anchor_len,
1767                       &error_fatal);
1768 
1769     if (smbios_anchor) {
1770         fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-tables",
1771                         smbios_tables, smbios_tables_len);
1772         fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-anchor",
1773                         smbios_anchor, smbios_anchor_len);
1774     }
1775 }
1776 
1777 static
virt_machine_done(Notifier * notifier,void * data)1778 void virt_machine_done(Notifier *notifier, void *data)
1779 {
1780     VirtMachineState *vms = container_of(notifier, VirtMachineState,
1781                                          machine_done);
1782     MachineState *ms = MACHINE(vms);
1783     ARMCPU *cpu = ARM_CPU(first_cpu);
1784     struct arm_boot_info *info = &vms->bootinfo;
1785     AddressSpace *as = arm_boot_address_space(cpu, info);
1786 
1787     cxl_hook_up_pxb_registers(vms->bus, &vms->cxl_devices_state,
1788                               &error_fatal);
1789 
1790     if (vms->cxl_devices_state.is_enabled) {
1791         cxl_fmws_link_targets(&error_fatal);
1792     }
1793     /*
1794      * If the user provided a dtb, we assume the dynamic sysbus nodes
1795      * already are integrated there. This corresponds to a use case where
1796      * the dynamic sysbus nodes are complex and their generation is not yet
1797      * supported. In that case the user can take charge of the guest dt
1798      * while qemu takes charge of the qom stuff.
1799      */
1800     if (info->dtb_filename == NULL) {
1801         platform_bus_add_all_fdt_nodes(ms->fdt, "/intc",
1802                                        vms->memmap[VIRT_PLATFORM_BUS].base,
1803                                        vms->memmap[VIRT_PLATFORM_BUS].size,
1804                                        vms->irqmap[VIRT_PLATFORM_BUS]);
1805     }
1806     if (arm_load_dtb(info->dtb_start, info, info->dtb_limit, as, ms, cpu) < 0) {
1807         exit(1);
1808     }
1809 
1810     pci_bus_add_fw_cfg_extra_pci_roots(vms->fw_cfg, vms->bus,
1811                                        &error_abort);
1812 
1813     virt_acpi_setup(vms);
1814     virt_build_smbios(vms);
1815 }
1816 
virt_cpu_mp_affinity(VirtMachineState * vms,int idx)1817 static uint64_t virt_cpu_mp_affinity(VirtMachineState *vms, int idx)
1818 {
1819     uint8_t clustersz;
1820 
1821     /*
1822      * Adjust MPIDR to make TCG consistent (with 64-bit KVM hosts)
1823      * and to improve SGI efficiency.
1824      */
1825     if (vms->gic_version == VIRT_GIC_VERSION_2) {
1826         clustersz = GIC_TARGETLIST_BITS;
1827     } else {
1828         clustersz = GICV3_TARGETLIST_BITS;
1829     }
1830 
1831     return arm_build_mp_affinity(idx, clustersz);
1832 }
1833 
virt_get_high_memmap_enabled(VirtMachineState * vms,int index)1834 static inline bool *virt_get_high_memmap_enabled(VirtMachineState *vms,
1835                                                  int index)
1836 {
1837     bool *enabled_array[] = {
1838         &vms->highmem_redists,
1839         &vms->highmem_cxl,
1840         &vms->highmem_ecam,
1841         &vms->highmem_mmio,
1842     };
1843 
1844     assert(ARRAY_SIZE(extended_memmap) - VIRT_LOWMEMMAP_LAST ==
1845            ARRAY_SIZE(enabled_array));
1846     assert(index - VIRT_LOWMEMMAP_LAST < ARRAY_SIZE(enabled_array));
1847 
1848     return enabled_array[index - VIRT_LOWMEMMAP_LAST];
1849 }
1850 
virt_set_high_memmap(VirtMachineState * vms,hwaddr base,int pa_bits)1851 static void virt_set_high_memmap(VirtMachineState *vms,
1852                                  hwaddr base, int pa_bits)
1853 {
1854     hwaddr region_base, region_size;
1855     bool *region_enabled, fits;
1856     int i;
1857 
1858     for (i = VIRT_LOWMEMMAP_LAST; i < ARRAY_SIZE(extended_memmap); i++) {
1859         region_enabled = virt_get_high_memmap_enabled(vms, i);
1860         region_base = ROUND_UP(base, extended_memmap[i].size);
1861         region_size = extended_memmap[i].size;
1862 
1863         vms->memmap[i].base = region_base;
1864         vms->memmap[i].size = region_size;
1865 
1866         /*
1867          * Check each device to see if it fits in the PA space,
1868          * moving highest_gpa as we go. For compatibility, move
1869          * highest_gpa for disabled fitting devices as well, if
1870          * the compact layout has been disabled.
1871          *
1872          * For each device that doesn't fit, disable it.
1873          */
1874         fits = (region_base + region_size) <= BIT_ULL(pa_bits);
1875         *region_enabled &= fits;
1876         if (vms->highmem_compact && !*region_enabled) {
1877             continue;
1878         }
1879 
1880         base = region_base + region_size;
1881         if (fits) {
1882             vms->highest_gpa = base - 1;
1883         }
1884     }
1885 }
1886 
virt_set_memmap(VirtMachineState * vms,int pa_bits)1887 static void virt_set_memmap(VirtMachineState *vms, int pa_bits)
1888 {
1889     MachineState *ms = MACHINE(vms);
1890     hwaddr base, device_memory_base, device_memory_size, memtop;
1891     int i;
1892 
1893     vms->memmap = extended_memmap;
1894 
1895     for (i = 0; i < ARRAY_SIZE(base_memmap); i++) {
1896         vms->memmap[i] = base_memmap[i];
1897     }
1898 
1899     if (ms->ram_slots > ACPI_MAX_RAM_SLOTS) {
1900         error_report("unsupported number of memory slots: %"PRIu64,
1901                      ms->ram_slots);
1902         exit(EXIT_FAILURE);
1903     }
1904 
1905     /*
1906      * !highmem is exactly the same as limiting the PA space to 32bit,
1907      * irrespective of the underlying capabilities of the HW.
1908      */
1909     if (!vms->highmem) {
1910         pa_bits = 32;
1911     }
1912 
1913     /*
1914      * We compute the base of the high IO region depending on the
1915      * amount of initial and device memory. The device memory start/size
1916      * is aligned on 1GiB. We never put the high IO region below 256GiB
1917      * so that if maxram_size is < 255GiB we keep the legacy memory map.
1918      * The device region size assumes 1GiB page max alignment per slot.
1919      */
1920     device_memory_base =
1921         ROUND_UP(vms->memmap[VIRT_MEM].base + ms->ram_size, GiB);
1922     device_memory_size = ms->maxram_size - ms->ram_size + ms->ram_slots * GiB;
1923 
1924     /* Base address of the high IO region */
1925     memtop = base = device_memory_base + ROUND_UP(device_memory_size, GiB);
1926     if (memtop > BIT_ULL(pa_bits)) {
1927         error_report("Addressing limited to %d bits, but memory exceeds it by %llu bytes",
1928                      pa_bits, memtop - BIT_ULL(pa_bits));
1929         exit(EXIT_FAILURE);
1930     }
1931     if (base < device_memory_base) {
1932         error_report("maxmem/slots too huge");
1933         exit(EXIT_FAILURE);
1934     }
1935     if (base < vms->memmap[VIRT_MEM].base + LEGACY_RAMLIMIT_BYTES) {
1936         base = vms->memmap[VIRT_MEM].base + LEGACY_RAMLIMIT_BYTES;
1937     }
1938 
1939     /* We know for sure that at least the memory fits in the PA space */
1940     vms->highest_gpa = memtop - 1;
1941 
1942     virt_set_high_memmap(vms, base, pa_bits);
1943 
1944     if (device_memory_size > 0) {
1945         machine_memory_devices_init(ms, device_memory_base, device_memory_size);
1946     }
1947     vms->highest_gpa = cxl_fmws_set_memmap(ROUND_UP(vms->highest_gpa + 1,
1948                                                     256 * MiB),
1949                                            BIT_ULL(pa_bits)) - 1;
1950 }
1951 
finalize_gic_version_do(const char * accel_name,VirtGICType gic_version,int gics_supported,unsigned int max_cpus)1952 static VirtGICType finalize_gic_version_do(const char *accel_name,
1953                                            VirtGICType gic_version,
1954                                            int gics_supported,
1955                                            unsigned int max_cpus)
1956 {
1957     /* Convert host/max/nosel to GIC version number */
1958     switch (gic_version) {
1959     case VIRT_GIC_VERSION_HOST:
1960         if (!kvm_enabled()) {
1961             error_report("gic-version=host requires KVM");
1962             exit(1);
1963         }
1964 
1965         /* For KVM, gic-version=host means gic-version=max */
1966         return finalize_gic_version_do(accel_name, VIRT_GIC_VERSION_MAX,
1967                                        gics_supported, max_cpus);
1968     case VIRT_GIC_VERSION_MAX:
1969         if (gics_supported & VIRT_GIC_VERSION_4_MASK) {
1970             gic_version = VIRT_GIC_VERSION_4;
1971         } else if (gics_supported & VIRT_GIC_VERSION_3_MASK) {
1972             gic_version = VIRT_GIC_VERSION_3;
1973         } else {
1974             gic_version = VIRT_GIC_VERSION_2;
1975         }
1976         break;
1977     case VIRT_GIC_VERSION_NOSEL:
1978         if ((gics_supported & VIRT_GIC_VERSION_2_MASK) &&
1979             max_cpus <= GIC_NCPU) {
1980             gic_version = VIRT_GIC_VERSION_2;
1981         } else if (gics_supported & VIRT_GIC_VERSION_3_MASK) {
1982             /*
1983              * in case the host does not support v2 emulation or
1984              * the end-user requested more than 8 VCPUs we now default
1985              * to v3. In any case defaulting to v2 would be broken.
1986              */
1987             gic_version = VIRT_GIC_VERSION_3;
1988         } else if (max_cpus > GIC_NCPU) {
1989             error_report("%s only supports GICv2 emulation but more than 8 "
1990                          "vcpus are requested", accel_name);
1991             exit(1);
1992         }
1993         break;
1994     case VIRT_GIC_VERSION_2:
1995     case VIRT_GIC_VERSION_3:
1996     case VIRT_GIC_VERSION_4:
1997         break;
1998     }
1999 
2000     /* Check chosen version is effectively supported */
2001     switch (gic_version) {
2002     case VIRT_GIC_VERSION_2:
2003         if (!(gics_supported & VIRT_GIC_VERSION_2_MASK)) {
2004             error_report("%s does not support GICv2 emulation", accel_name);
2005             exit(1);
2006         }
2007         break;
2008     case VIRT_GIC_VERSION_3:
2009         if (!(gics_supported & VIRT_GIC_VERSION_3_MASK)) {
2010             error_report("%s does not support GICv3 emulation", accel_name);
2011             exit(1);
2012         }
2013         break;
2014     case VIRT_GIC_VERSION_4:
2015         if (!(gics_supported & VIRT_GIC_VERSION_4_MASK)) {
2016             error_report("%s does not support GICv4 emulation, is virtualization=on?",
2017                          accel_name);
2018             exit(1);
2019         }
2020         break;
2021     default:
2022         error_report("logic error in finalize_gic_version");
2023         exit(1);
2024         break;
2025     }
2026 
2027     return gic_version;
2028 }
2029 
2030 /*
2031  * finalize_gic_version - Determines the final gic_version
2032  * according to the gic-version property
2033  *
2034  * Default GIC type is v2
2035  */
finalize_gic_version(VirtMachineState * vms)2036 static void finalize_gic_version(VirtMachineState *vms)
2037 {
2038     const char *accel_name = current_accel_name();
2039     unsigned int max_cpus = MACHINE(vms)->smp.max_cpus;
2040     int gics_supported = 0;
2041 
2042     /* Determine which GIC versions the current environment supports */
2043     if (kvm_enabled() && kvm_irqchip_in_kernel()) {
2044         int probe_bitmap = kvm_arm_vgic_probe();
2045 
2046         if (!probe_bitmap) {
2047             error_report("Unable to determine GIC version supported by host");
2048             exit(1);
2049         }
2050 
2051         if (probe_bitmap & KVM_ARM_VGIC_V2) {
2052             gics_supported |= VIRT_GIC_VERSION_2_MASK;
2053         }
2054         if (probe_bitmap & KVM_ARM_VGIC_V3) {
2055             gics_supported |= VIRT_GIC_VERSION_3_MASK;
2056         }
2057     } else if (kvm_enabled() && !kvm_irqchip_in_kernel()) {
2058         /* KVM w/o kernel irqchip can only deal with GICv2 */
2059         gics_supported |= VIRT_GIC_VERSION_2_MASK;
2060         accel_name = "KVM with kernel-irqchip=off";
2061     } else if (tcg_enabled() || hvf_enabled() || qtest_enabled())  {
2062         gics_supported |= VIRT_GIC_VERSION_2_MASK;
2063         if (module_object_class_by_name("arm-gicv3")) {
2064             gics_supported |= VIRT_GIC_VERSION_3_MASK;
2065             if (vms->virt) {
2066                 /* GICv4 only makes sense if CPU has EL2 */
2067                 gics_supported |= VIRT_GIC_VERSION_4_MASK;
2068             }
2069         }
2070     } else {
2071         error_report("Unsupported accelerator, can not determine GIC support");
2072         exit(1);
2073     }
2074 
2075     /*
2076      * Then convert helpers like host/max to concrete GIC versions and ensure
2077      * the desired version is supported
2078      */
2079     vms->gic_version = finalize_gic_version_do(accel_name, vms->gic_version,
2080                                                gics_supported, max_cpus);
2081 }
2082 
2083 /*
2084  * virt_post_cpus_gic_realized() must be called after the CPUs and
2085  * the GIC have both been realized.
2086  */
virt_post_cpus_gic_realized(VirtMachineState * vms,MemoryRegion * sysmem)2087 static void virt_post_cpus_gic_realized(VirtMachineState *vms,
2088                                         MemoryRegion *sysmem)
2089 {
2090     int max_cpus = MACHINE(vms)->smp.max_cpus;
2091     bool aarch64, pmu, steal_time;
2092     CPUState *cpu;
2093 
2094     aarch64 = object_property_get_bool(OBJECT(first_cpu), "aarch64", NULL);
2095     pmu = object_property_get_bool(OBJECT(first_cpu), "pmu", NULL);
2096     steal_time = object_property_get_bool(OBJECT(first_cpu),
2097                                           "kvm-steal-time", NULL);
2098 
2099     if (kvm_enabled()) {
2100         hwaddr pvtime_reg_base = vms->memmap[VIRT_PVTIME].base;
2101         hwaddr pvtime_reg_size = vms->memmap[VIRT_PVTIME].size;
2102 
2103         if (steal_time) {
2104             MemoryRegion *pvtime = g_new(MemoryRegion, 1);
2105             hwaddr pvtime_size = max_cpus * PVTIME_SIZE_PER_CPU;
2106 
2107             /* The memory region size must be a multiple of host page size. */
2108             pvtime_size = REAL_HOST_PAGE_ALIGN(pvtime_size);
2109 
2110             if (pvtime_size > pvtime_reg_size) {
2111                 error_report("pvtime requires a %" HWADDR_PRId
2112                              " byte memory region for %d CPUs,"
2113                              " but only %" HWADDR_PRId " has been reserved",
2114                              pvtime_size, max_cpus, pvtime_reg_size);
2115                 exit(1);
2116             }
2117 
2118             memory_region_init_ram(pvtime, NULL, "pvtime", pvtime_size, NULL);
2119             memory_region_add_subregion(sysmem, pvtime_reg_base, pvtime);
2120         }
2121         if (!aarch64 && vms->virt) {
2122             error_report("KVM does not support EL2 on an AArch32 vCPU");
2123             exit(1);
2124         }
2125 
2126         CPU_FOREACH(cpu) {
2127             if (pmu) {
2128                 assert(arm_feature(&ARM_CPU(cpu)->env, ARM_FEATURE_PMU));
2129                 if (kvm_irqchip_in_kernel()) {
2130                     kvm_arm_pmu_set_irq(ARM_CPU(cpu), VIRTUAL_PMU_IRQ);
2131                 }
2132                 kvm_arm_pmu_init(ARM_CPU(cpu));
2133             }
2134             if (steal_time) {
2135                 kvm_arm_pvtime_init(ARM_CPU(cpu), pvtime_reg_base
2136                                                   + cpu->cpu_index
2137                                                     * PVTIME_SIZE_PER_CPU);
2138             }
2139         }
2140     } else {
2141         if (aarch64 && vms->highmem) {
2142             int requested_pa_size = 64 - clz64(vms->highest_gpa);
2143             int pamax = arm_pamax(ARM_CPU(first_cpu));
2144 
2145             if (pamax < requested_pa_size) {
2146                 error_report("VCPU supports less PA bits (%d) than "
2147                              "requested by the memory map (%d)",
2148                              pamax, requested_pa_size);
2149                 exit(1);
2150             }
2151         }
2152     }
2153 }
2154 
machvirt_init(MachineState * machine)2155 static void machvirt_init(MachineState *machine)
2156 {
2157     VirtMachineState *vms = VIRT_MACHINE(machine);
2158     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(machine);
2159     MachineClass *mc = MACHINE_GET_CLASS(machine);
2160     const CPUArchIdList *possible_cpus;
2161     MemoryRegion *sysmem = get_system_memory();
2162     MemoryRegion *secure_sysmem = NULL;
2163     MemoryRegion *tag_sysmem = NULL;
2164     MemoryRegion *secure_tag_sysmem = NULL;
2165     int n, virt_max_cpus;
2166     bool firmware_loaded;
2167     bool aarch64 = true;
2168     bool has_ged = !vmc->no_ged;
2169     unsigned int smp_cpus = machine->smp.cpus;
2170     unsigned int max_cpus = machine->smp.max_cpus;
2171 
2172     possible_cpus = mc->possible_cpu_arch_ids(machine);
2173 
2174     /*
2175      * In accelerated mode, the memory map is computed earlier in kvm_type()
2176      * for Linux, or hvf_get_physical_address_range() for macOS to create a
2177      * VM with the right number of IPA bits.
2178      */
2179     if (!vms->memmap) {
2180         Object *cpuobj;
2181         ARMCPU *armcpu;
2182         int pa_bits;
2183 
2184         /*
2185          * Instantiate a temporary CPU object to find out about what
2186          * we are about to deal with. Once this is done, get rid of
2187          * the object.
2188          */
2189         cpuobj = object_new(possible_cpus->cpus[0].type);
2190         armcpu = ARM_CPU(cpuobj);
2191 
2192         pa_bits = arm_pamax(armcpu);
2193 
2194         object_unref(cpuobj);
2195 
2196         virt_set_memmap(vms, pa_bits);
2197     }
2198 
2199     /* We can probe only here because during property set
2200      * KVM is not available yet
2201      */
2202     finalize_gic_version(vms);
2203 
2204     if (vms->secure) {
2205         /*
2206          * The Secure view of the world is the same as the NonSecure,
2207          * but with a few extra devices. Create it as a container region
2208          * containing the system memory at low priority; any secure-only
2209          * devices go in at higher priority and take precedence.
2210          */
2211         secure_sysmem = g_new(MemoryRegion, 1);
2212         memory_region_init(secure_sysmem, OBJECT(machine), "secure-memory",
2213                            UINT64_MAX);
2214         memory_region_add_subregion_overlap(secure_sysmem, 0, sysmem, -1);
2215     }
2216 
2217     firmware_loaded = virt_firmware_init(vms, sysmem,
2218                                          secure_sysmem ?: sysmem);
2219 
2220     /* If we have an EL3 boot ROM then the assumption is that it will
2221      * implement PSCI itself, so disable QEMU's internal implementation
2222      * so it doesn't get in the way. Instead of starting secondary
2223      * CPUs in PSCI powerdown state we will start them all running and
2224      * let the boot ROM sort them out.
2225      * The usual case is that we do use QEMU's PSCI implementation;
2226      * if the guest has EL2 then we will use SMC as the conduit,
2227      * and otherwise we will use HVC (for backwards compatibility and
2228      * because if we're using KVM then we must use HVC).
2229      */
2230     if (vms->secure && firmware_loaded) {
2231         vms->psci_conduit = QEMU_PSCI_CONDUIT_DISABLED;
2232     } else if (vms->virt) {
2233         vms->psci_conduit = QEMU_PSCI_CONDUIT_SMC;
2234     } else {
2235         vms->psci_conduit = QEMU_PSCI_CONDUIT_HVC;
2236     }
2237 
2238     /*
2239      * The maximum number of CPUs depends on the GIC version, or on how
2240      * many redistributors we can fit into the memory map (which in turn
2241      * depends on whether this is a GICv3 or v4).
2242      */
2243     if (vms->gic_version == VIRT_GIC_VERSION_2) {
2244         virt_max_cpus = GIC_NCPU;
2245     } else {
2246         virt_max_cpus = virt_redist_capacity(vms, VIRT_GIC_REDIST);
2247         if (vms->highmem_redists) {
2248             virt_max_cpus += virt_redist_capacity(vms, VIRT_HIGH_GIC_REDIST2);
2249         }
2250     }
2251 
2252     if (max_cpus > virt_max_cpus) {
2253         error_report("Number of SMP CPUs requested (%d) exceeds max CPUs "
2254                      "supported by machine 'mach-virt' (%d)",
2255                      max_cpus, virt_max_cpus);
2256         if (vms->gic_version != VIRT_GIC_VERSION_2 && !vms->highmem_redists) {
2257             error_printf("Try 'highmem-redists=on' for more CPUs\n");
2258         }
2259 
2260         exit(1);
2261     }
2262 
2263     if (vms->secure && !tcg_enabled() && !qtest_enabled()) {
2264         error_report("mach-virt: %s does not support providing "
2265                      "Security extensions (TrustZone) to the guest CPU",
2266                      current_accel_name());
2267         exit(1);
2268     }
2269 
2270     if (vms->virt && kvm_enabled() && !kvm_arm_el2_supported()) {
2271         error_report("mach-virt: host kernel KVM does not support providing "
2272                      "Virtualization extensions to the guest CPU");
2273         exit(1);
2274     }
2275 
2276     if (vms->virt && !kvm_enabled() && !tcg_enabled() && !qtest_enabled()) {
2277         error_report("mach-virt: %s does not support providing "
2278                      "Virtualization extensions to the guest CPU",
2279                      current_accel_name());
2280         exit(1);
2281     }
2282 
2283     if (vms->mte && hvf_enabled()) {
2284         error_report("mach-virt: %s does not support providing "
2285                      "MTE to the guest CPU",
2286                      current_accel_name());
2287         exit(1);
2288     }
2289 
2290     create_fdt(vms);
2291 
2292     assert(possible_cpus->len == max_cpus);
2293     for (n = 0; n < possible_cpus->len; n++) {
2294         Object *cpuobj;
2295         CPUState *cs;
2296 
2297         if (n >= smp_cpus) {
2298             break;
2299         }
2300 
2301         cpuobj = object_new(possible_cpus->cpus[n].type);
2302         object_property_set_int(cpuobj, "mp-affinity",
2303                                 possible_cpus->cpus[n].arch_id, NULL);
2304 
2305         cs = CPU(cpuobj);
2306         cs->cpu_index = n;
2307 
2308         numa_cpu_pre_plug(&possible_cpus->cpus[cs->cpu_index], DEVICE(cpuobj),
2309                           &error_fatal);
2310 
2311         aarch64 &= object_property_get_bool(cpuobj, "aarch64", NULL);
2312 
2313         if (!vms->secure) {
2314             object_property_set_bool(cpuobj, "has_el3", false, NULL);
2315         }
2316 
2317         if (!vms->virt && object_property_find(cpuobj, "has_el2")) {
2318             object_property_set_bool(cpuobj, "has_el2", false, NULL);
2319         }
2320 
2321         if (vmc->kvm_no_adjvtime &&
2322             object_property_find(cpuobj, "kvm-no-adjvtime")) {
2323             object_property_set_bool(cpuobj, "kvm-no-adjvtime", true, NULL);
2324         }
2325 
2326         if (vmc->no_kvm_steal_time &&
2327             object_property_find(cpuobj, "kvm-steal-time")) {
2328             object_property_set_bool(cpuobj, "kvm-steal-time", false, NULL);
2329         }
2330 
2331         if (vmc->no_tcg_lpa2 && object_property_find(cpuobj, "lpa2")) {
2332             object_property_set_bool(cpuobj, "lpa2", false, NULL);
2333         }
2334 
2335         if (object_property_find(cpuobj, "reset-cbar")) {
2336             object_property_set_int(cpuobj, "reset-cbar",
2337                                     vms->memmap[VIRT_CPUPERIPHS].base,
2338                                     &error_abort);
2339         }
2340 
2341         object_property_set_link(cpuobj, "memory", OBJECT(sysmem),
2342                                  &error_abort);
2343         if (vms->secure) {
2344             object_property_set_link(cpuobj, "secure-memory",
2345                                      OBJECT(secure_sysmem), &error_abort);
2346         }
2347 
2348         if (vms->mte) {
2349             if (tcg_enabled()) {
2350                 /* Create the memory region only once, but link to all cpus. */
2351                 if (!tag_sysmem) {
2352                     /*
2353                      * The property exists only if MemTag is supported.
2354                      * If it is, we must allocate the ram to back that up.
2355                      */
2356                     if (!object_property_find(cpuobj, "tag-memory")) {
2357                         error_report("MTE requested, but not supported "
2358                                      "by the guest CPU");
2359                         exit(1);
2360                     }
2361 
2362                     tag_sysmem = g_new(MemoryRegion, 1);
2363                     memory_region_init(tag_sysmem, OBJECT(machine),
2364                                        "tag-memory", UINT64_MAX / 32);
2365 
2366                     if (vms->secure) {
2367                         secure_tag_sysmem = g_new(MemoryRegion, 1);
2368                         memory_region_init(secure_tag_sysmem, OBJECT(machine),
2369                                            "secure-tag-memory",
2370                                            UINT64_MAX / 32);
2371 
2372                         /* As with ram, secure-tag takes precedence over tag. */
2373                         memory_region_add_subregion_overlap(secure_tag_sysmem,
2374                                                             0, tag_sysmem, -1);
2375                     }
2376                 }
2377 
2378                 object_property_set_link(cpuobj, "tag-memory",
2379                                          OBJECT(tag_sysmem), &error_abort);
2380                 if (vms->secure) {
2381                     object_property_set_link(cpuobj, "secure-tag-memory",
2382                                              OBJECT(secure_tag_sysmem),
2383                                              &error_abort);
2384                 }
2385             } else if (kvm_enabled()) {
2386                 if (!kvm_arm_mte_supported()) {
2387                     error_report("MTE requested, but not supported by KVM");
2388                     exit(1);
2389                 }
2390                 kvm_arm_enable_mte(cpuobj, &error_abort);
2391             } else {
2392                     error_report("MTE requested, but not supported ");
2393                     exit(1);
2394             }
2395         }
2396 
2397         qdev_realize(DEVICE(cpuobj), NULL, &error_fatal);
2398         object_unref(cpuobj);
2399     }
2400 
2401     /* Now we've created the CPUs we can see if they have the hypvirt timer */
2402     vms->ns_el2_virt_timer_irq = ns_el2_virt_timer_present() &&
2403         !vmc->no_ns_el2_virt_timer_irq;
2404 
2405     fdt_add_timer_nodes(vms);
2406     fdt_add_cpu_nodes(vms);
2407 
2408     memory_region_add_subregion(sysmem, vms->memmap[VIRT_MEM].base,
2409                                 machine->ram);
2410 
2411     cxl_fmws_update_mmio();
2412 
2413     virt_flash_fdt(vms, sysmem, secure_sysmem ?: sysmem);
2414 
2415     create_gic(vms, sysmem);
2416 
2417     virt_post_cpus_gic_realized(vms, sysmem);
2418 
2419     fdt_add_pmu_nodes(vms);
2420 
2421     /*
2422      * The first UART always exists. If the security extensions are
2423      * enabled, the second UART also always exists. Otherwise, it only exists
2424      * if a backend is configured explicitly via '-serial <backend>'.
2425      * This avoids potentially breaking existing user setups that expect
2426      * only one NonSecure UART to be present (for instance, older EDK2
2427      * binaries).
2428      *
2429      * The nodes end up in the DTB in reverse order of creation, so we must
2430      * create UART0 last to ensure it appears as the first node in the DTB,
2431      * for compatibility with guest software that just iterates through the
2432      * DTB to find the first UART, as older versions of EDK2 do.
2433      * DTB readers that follow the spec, as Linux does, should honour the
2434      * aliases node information and /chosen/stdout-path regardless of
2435      * the order that nodes appear in the DTB.
2436      *
2437      * For similar back-compatibility reasons, if UART1 is the secure UART
2438      * we create it second (and so it appears first in the DTB), because
2439      * that's what QEMU has always done.
2440      */
2441     if (!vms->secure) {
2442         Chardev *serial1 = serial_hd(1);
2443 
2444         if (serial1) {
2445             vms->second_ns_uart_present = true;
2446             create_uart(vms, VIRT_UART1, sysmem, serial1, false);
2447         }
2448     }
2449     create_uart(vms, VIRT_UART0, sysmem, serial_hd(0), false);
2450     if (vms->secure) {
2451         create_uart(vms, VIRT_UART1, secure_sysmem, serial_hd(1), true);
2452     }
2453 
2454     if (vms->secure) {
2455         create_secure_ram(vms, secure_sysmem, secure_tag_sysmem);
2456     }
2457 
2458     if (tag_sysmem) {
2459         create_tag_ram(tag_sysmem, vms->memmap[VIRT_MEM].base,
2460                        machine->ram_size, "mach-virt.tag");
2461     }
2462 
2463     vms->highmem_ecam &= (!firmware_loaded || aarch64);
2464 
2465     create_rtc(vms);
2466 
2467     create_pcie(vms);
2468     create_cxl_host_reg_region(vms);
2469 
2470     if (has_ged && aarch64 && firmware_loaded && virt_is_acpi_enabled(vms)) {
2471         vms->acpi_dev = create_acpi_ged(vms);
2472     } else {
2473         create_gpio_devices(vms, VIRT_GPIO, sysmem);
2474     }
2475 
2476     if (vms->secure && !vmc->no_secure_gpio) {
2477         create_gpio_devices(vms, VIRT_SECURE_GPIO, secure_sysmem);
2478     }
2479 
2480      /* connect powerdown request */
2481      vms->powerdown_notifier.notify = virt_powerdown_req;
2482      qemu_register_powerdown_notifier(&vms->powerdown_notifier);
2483 
2484     /* Create mmio transports, so the user can create virtio backends
2485      * (which will be automatically plugged in to the transports). If
2486      * no backend is created the transport will just sit harmlessly idle.
2487      */
2488     create_virtio_devices(vms);
2489 
2490     vms->fw_cfg = create_fw_cfg(vms, &address_space_memory);
2491     rom_set_fw(vms->fw_cfg);
2492 
2493     create_platform_bus(vms);
2494 
2495     if (machine->nvdimms_state->is_enabled) {
2496         const struct AcpiGenericAddress arm_virt_nvdimm_acpi_dsmio = {
2497             .space_id = AML_AS_SYSTEM_MEMORY,
2498             .address = vms->memmap[VIRT_NVDIMM_ACPI].base,
2499             .bit_width = NVDIMM_ACPI_IO_LEN << 3
2500         };
2501 
2502         nvdimm_init_acpi_state(machine->nvdimms_state, sysmem,
2503                                arm_virt_nvdimm_acpi_dsmio,
2504                                vms->fw_cfg, OBJECT(vms));
2505     }
2506 
2507     vms->bootinfo.ram_size = machine->ram_size;
2508     vms->bootinfo.board_id = -1;
2509     vms->bootinfo.loader_start = vms->memmap[VIRT_MEM].base;
2510     vms->bootinfo.get_dtb = machvirt_dtb;
2511     vms->bootinfo.skip_dtb_autoload = true;
2512     vms->bootinfo.firmware_loaded = firmware_loaded;
2513     vms->bootinfo.psci_conduit = vms->psci_conduit;
2514     arm_load_kernel(ARM_CPU(first_cpu), machine, &vms->bootinfo);
2515 
2516     vms->machine_done.notify = virt_machine_done;
2517     qemu_add_machine_init_done_notifier(&vms->machine_done);
2518 }
2519 
virt_get_secure(Object * obj,Error ** errp)2520 static bool virt_get_secure(Object *obj, Error **errp)
2521 {
2522     VirtMachineState *vms = VIRT_MACHINE(obj);
2523 
2524     return vms->secure;
2525 }
2526 
virt_set_secure(Object * obj,bool value,Error ** errp)2527 static void virt_set_secure(Object *obj, bool value, Error **errp)
2528 {
2529     VirtMachineState *vms = VIRT_MACHINE(obj);
2530 
2531     vms->secure = value;
2532 }
2533 
virt_get_virt(Object * obj,Error ** errp)2534 static bool virt_get_virt(Object *obj, Error **errp)
2535 {
2536     VirtMachineState *vms = VIRT_MACHINE(obj);
2537 
2538     return vms->virt;
2539 }
2540 
virt_set_virt(Object * obj,bool value,Error ** errp)2541 static void virt_set_virt(Object *obj, bool value, Error **errp)
2542 {
2543     VirtMachineState *vms = VIRT_MACHINE(obj);
2544 
2545     vms->virt = value;
2546 }
2547 
virt_get_highmem(Object * obj,Error ** errp)2548 static bool virt_get_highmem(Object *obj, Error **errp)
2549 {
2550     VirtMachineState *vms = VIRT_MACHINE(obj);
2551 
2552     return vms->highmem;
2553 }
2554 
virt_set_highmem(Object * obj,bool value,Error ** errp)2555 static void virt_set_highmem(Object *obj, bool value, Error **errp)
2556 {
2557     VirtMachineState *vms = VIRT_MACHINE(obj);
2558 
2559     vms->highmem = value;
2560 }
2561 
virt_get_compact_highmem(Object * obj,Error ** errp)2562 static bool virt_get_compact_highmem(Object *obj, Error **errp)
2563 {
2564     VirtMachineState *vms = VIRT_MACHINE(obj);
2565 
2566     return vms->highmem_compact;
2567 }
2568 
virt_set_compact_highmem(Object * obj,bool value,Error ** errp)2569 static void virt_set_compact_highmem(Object *obj, bool value, Error **errp)
2570 {
2571     VirtMachineState *vms = VIRT_MACHINE(obj);
2572 
2573     vms->highmem_compact = value;
2574 }
2575 
virt_get_highmem_redists(Object * obj,Error ** errp)2576 static bool virt_get_highmem_redists(Object *obj, Error **errp)
2577 {
2578     VirtMachineState *vms = VIRT_MACHINE(obj);
2579 
2580     return vms->highmem_redists;
2581 }
2582 
virt_set_highmem_redists(Object * obj,bool value,Error ** errp)2583 static void virt_set_highmem_redists(Object *obj, bool value, Error **errp)
2584 {
2585     VirtMachineState *vms = VIRT_MACHINE(obj);
2586 
2587     vms->highmem_redists = value;
2588 }
2589 
virt_get_highmem_ecam(Object * obj,Error ** errp)2590 static bool virt_get_highmem_ecam(Object *obj, Error **errp)
2591 {
2592     VirtMachineState *vms = VIRT_MACHINE(obj);
2593 
2594     return vms->highmem_ecam;
2595 }
2596 
virt_set_highmem_ecam(Object * obj,bool value,Error ** errp)2597 static void virt_set_highmem_ecam(Object *obj, bool value, Error **errp)
2598 {
2599     VirtMachineState *vms = VIRT_MACHINE(obj);
2600 
2601     vms->highmem_ecam = value;
2602 }
2603 
virt_get_highmem_mmio(Object * obj,Error ** errp)2604 static bool virt_get_highmem_mmio(Object *obj, Error **errp)
2605 {
2606     VirtMachineState *vms = VIRT_MACHINE(obj);
2607 
2608     return vms->highmem_mmio;
2609 }
2610 
virt_set_highmem_mmio(Object * obj,bool value,Error ** errp)2611 static void virt_set_highmem_mmio(Object *obj, bool value, Error **errp)
2612 {
2613     VirtMachineState *vms = VIRT_MACHINE(obj);
2614 
2615     vms->highmem_mmio = value;
2616 }
2617 
virt_get_highmem_mmio_size(Object * obj,Visitor * v,const char * name,void * opaque,Error ** errp)2618 static void virt_get_highmem_mmio_size(Object *obj, Visitor *v,
2619                                        const char *name, void *opaque,
2620                                        Error **errp)
2621 {
2622     uint64_t size = extended_memmap[VIRT_HIGH_PCIE_MMIO].size;
2623 
2624     visit_type_size(v, name, &size, errp);
2625 }
2626 
virt_set_highmem_mmio_size(Object * obj,Visitor * v,const char * name,void * opaque,Error ** errp)2627 static void virt_set_highmem_mmio_size(Object *obj, Visitor *v,
2628                                        const char *name, void *opaque,
2629                                        Error **errp)
2630 {
2631     uint64_t size;
2632 
2633     if (!visit_type_size(v, name, &size, errp)) {
2634         return;
2635     }
2636 
2637     if (!is_power_of_2(size)) {
2638         error_setg(errp, "highmem-mmio-size is not a power of 2");
2639         return;
2640     }
2641 
2642     if (size < DEFAULT_HIGH_PCIE_MMIO_SIZE) {
2643         char *sz = size_to_str(DEFAULT_HIGH_PCIE_MMIO_SIZE);
2644         error_setg(errp, "highmem-mmio-size cannot be set to a lower value "
2645                          "than the default (%s)", sz);
2646         g_free(sz);
2647         return;
2648     }
2649 
2650     extended_memmap[VIRT_HIGH_PCIE_MMIO].size = size;
2651 }
2652 
virt_get_its(Object * obj,Error ** errp)2653 static bool virt_get_its(Object *obj, Error **errp)
2654 {
2655     VirtMachineState *vms = VIRT_MACHINE(obj);
2656 
2657     return vms->its;
2658 }
2659 
virt_set_its(Object * obj,bool value,Error ** errp)2660 static void virt_set_its(Object *obj, bool value, Error **errp)
2661 {
2662     VirtMachineState *vms = VIRT_MACHINE(obj);
2663 
2664     vms->its = value;
2665 }
2666 
virt_get_dtb_randomness(Object * obj,Error ** errp)2667 static bool virt_get_dtb_randomness(Object *obj, Error **errp)
2668 {
2669     VirtMachineState *vms = VIRT_MACHINE(obj);
2670 
2671     return vms->dtb_randomness;
2672 }
2673 
virt_set_dtb_randomness(Object * obj,bool value,Error ** errp)2674 static void virt_set_dtb_randomness(Object *obj, bool value, Error **errp)
2675 {
2676     VirtMachineState *vms = VIRT_MACHINE(obj);
2677 
2678     vms->dtb_randomness = value;
2679 }
2680 
virt_get_oem_id(Object * obj,Error ** errp)2681 static char *virt_get_oem_id(Object *obj, Error **errp)
2682 {
2683     VirtMachineState *vms = VIRT_MACHINE(obj);
2684 
2685     return g_strdup(vms->oem_id);
2686 }
2687 
virt_set_oem_id(Object * obj,const char * value,Error ** errp)2688 static void virt_set_oem_id(Object *obj, const char *value, Error **errp)
2689 {
2690     VirtMachineState *vms = VIRT_MACHINE(obj);
2691     size_t len = strlen(value);
2692 
2693     if (len > 6) {
2694         error_setg(errp,
2695                    "User specified oem-id value is bigger than 6 bytes in size");
2696         return;
2697     }
2698 
2699     strncpy(vms->oem_id, value, 6);
2700 }
2701 
virt_get_oem_table_id(Object * obj,Error ** errp)2702 static char *virt_get_oem_table_id(Object *obj, Error **errp)
2703 {
2704     VirtMachineState *vms = VIRT_MACHINE(obj);
2705 
2706     return g_strdup(vms->oem_table_id);
2707 }
2708 
virt_set_oem_table_id(Object * obj,const char * value,Error ** errp)2709 static void virt_set_oem_table_id(Object *obj, const char *value,
2710                                   Error **errp)
2711 {
2712     VirtMachineState *vms = VIRT_MACHINE(obj);
2713     size_t len = strlen(value);
2714 
2715     if (len > 8) {
2716         error_setg(errp,
2717                    "User specified oem-table-id value is bigger than 8 bytes in size");
2718         return;
2719     }
2720     strncpy(vms->oem_table_id, value, 8);
2721 }
2722 
2723 
virt_is_acpi_enabled(VirtMachineState * vms)2724 bool virt_is_acpi_enabled(VirtMachineState *vms)
2725 {
2726     if (vms->acpi == ON_OFF_AUTO_OFF) {
2727         return false;
2728     }
2729     return true;
2730 }
2731 
virt_get_acpi(Object * obj,Visitor * v,const char * name,void * opaque,Error ** errp)2732 static void virt_get_acpi(Object *obj, Visitor *v, const char *name,
2733                           void *opaque, Error **errp)
2734 {
2735     VirtMachineState *vms = VIRT_MACHINE(obj);
2736     OnOffAuto acpi = vms->acpi;
2737 
2738     visit_type_OnOffAuto(v, name, &acpi, errp);
2739 }
2740 
virt_set_acpi(Object * obj,Visitor * v,const char * name,void * opaque,Error ** errp)2741 static void virt_set_acpi(Object *obj, Visitor *v, const char *name,
2742                           void *opaque, Error **errp)
2743 {
2744     VirtMachineState *vms = VIRT_MACHINE(obj);
2745 
2746     visit_type_OnOffAuto(v, name, &vms->acpi, errp);
2747 }
2748 
virt_get_ras(Object * obj,Error ** errp)2749 static bool virt_get_ras(Object *obj, Error **errp)
2750 {
2751     VirtMachineState *vms = VIRT_MACHINE(obj);
2752 
2753     return vms->ras;
2754 }
2755 
virt_set_ras(Object * obj,bool value,Error ** errp)2756 static void virt_set_ras(Object *obj, bool value, Error **errp)
2757 {
2758     VirtMachineState *vms = VIRT_MACHINE(obj);
2759 
2760     vms->ras = value;
2761 }
2762 
virt_get_mte(Object * obj,Error ** errp)2763 static bool virt_get_mte(Object *obj, Error **errp)
2764 {
2765     VirtMachineState *vms = VIRT_MACHINE(obj);
2766 
2767     return vms->mte;
2768 }
2769 
virt_set_mte(Object * obj,bool value,Error ** errp)2770 static void virt_set_mte(Object *obj, bool value, Error **errp)
2771 {
2772     VirtMachineState *vms = VIRT_MACHINE(obj);
2773 
2774     vms->mte = value;
2775 }
2776 
virt_get_gic_version(Object * obj,Error ** errp)2777 static char *virt_get_gic_version(Object *obj, Error **errp)
2778 {
2779     VirtMachineState *vms = VIRT_MACHINE(obj);
2780     const char *val;
2781 
2782     switch (vms->gic_version) {
2783     case VIRT_GIC_VERSION_4:
2784         val = "4";
2785         break;
2786     case VIRT_GIC_VERSION_3:
2787         val = "3";
2788         break;
2789     default:
2790         val = "2";
2791         break;
2792     }
2793     return g_strdup(val);
2794 }
2795 
virt_set_gic_version(Object * obj,const char * value,Error ** errp)2796 static void virt_set_gic_version(Object *obj, const char *value, Error **errp)
2797 {
2798     VirtMachineState *vms = VIRT_MACHINE(obj);
2799 
2800     if (!strcmp(value, "4")) {
2801         vms->gic_version = VIRT_GIC_VERSION_4;
2802     } else if (!strcmp(value, "3")) {
2803         vms->gic_version = VIRT_GIC_VERSION_3;
2804     } else if (!strcmp(value, "2")) {
2805         vms->gic_version = VIRT_GIC_VERSION_2;
2806     } else if (!strcmp(value, "host")) {
2807         vms->gic_version = VIRT_GIC_VERSION_HOST; /* Will probe later */
2808     } else if (!strcmp(value, "max")) {
2809         vms->gic_version = VIRT_GIC_VERSION_MAX; /* Will probe later */
2810     } else {
2811         error_setg(errp, "Invalid gic-version value");
2812         error_append_hint(errp, "Valid values are 3, 2, host, max.\n");
2813     }
2814 }
2815 
virt_get_iommu(Object * obj,Error ** errp)2816 static char *virt_get_iommu(Object *obj, Error **errp)
2817 {
2818     VirtMachineState *vms = VIRT_MACHINE(obj);
2819 
2820     switch (vms->iommu) {
2821     case VIRT_IOMMU_NONE:
2822         return g_strdup("none");
2823     case VIRT_IOMMU_SMMUV3:
2824         return g_strdup("smmuv3");
2825     default:
2826         g_assert_not_reached();
2827     }
2828 }
2829 
virt_set_iommu(Object * obj,const char * value,Error ** errp)2830 static void virt_set_iommu(Object *obj, const char *value, Error **errp)
2831 {
2832     VirtMachineState *vms = VIRT_MACHINE(obj);
2833 
2834     if (!strcmp(value, "smmuv3")) {
2835         vms->iommu = VIRT_IOMMU_SMMUV3;
2836     } else if (!strcmp(value, "none")) {
2837         vms->iommu = VIRT_IOMMU_NONE;
2838     } else {
2839         error_setg(errp, "Invalid iommu value");
2840         error_append_hint(errp, "Valid values are none, smmuv3.\n");
2841     }
2842 }
2843 
virt_get_default_bus_bypass_iommu(Object * obj,Error ** errp)2844 static bool virt_get_default_bus_bypass_iommu(Object *obj, Error **errp)
2845 {
2846     VirtMachineState *vms = VIRT_MACHINE(obj);
2847 
2848     return vms->default_bus_bypass_iommu;
2849 }
2850 
virt_set_default_bus_bypass_iommu(Object * obj,bool value,Error ** errp)2851 static void virt_set_default_bus_bypass_iommu(Object *obj, bool value,
2852                                               Error **errp)
2853 {
2854     VirtMachineState *vms = VIRT_MACHINE(obj);
2855 
2856     vms->default_bus_bypass_iommu = value;
2857 }
2858 
2859 static CpuInstanceProperties
virt_cpu_index_to_props(MachineState * ms,unsigned cpu_index)2860 virt_cpu_index_to_props(MachineState *ms, unsigned cpu_index)
2861 {
2862     MachineClass *mc = MACHINE_GET_CLASS(ms);
2863     const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms);
2864 
2865     assert(cpu_index < possible_cpus->len);
2866     return possible_cpus->cpus[cpu_index].props;
2867 }
2868 
virt_get_default_cpu_node_id(const MachineState * ms,int idx)2869 static int64_t virt_get_default_cpu_node_id(const MachineState *ms, int idx)
2870 {
2871     int64_t socket_id = ms->possible_cpus->cpus[idx].props.socket_id;
2872 
2873     return socket_id % ms->numa_state->num_nodes;
2874 }
2875 
virt_possible_cpu_arch_ids(MachineState * ms)2876 static const CPUArchIdList *virt_possible_cpu_arch_ids(MachineState *ms)
2877 {
2878     int n;
2879     unsigned int max_cpus = ms->smp.max_cpus;
2880     VirtMachineState *vms = VIRT_MACHINE(ms);
2881     MachineClass *mc = MACHINE_GET_CLASS(vms);
2882 
2883     if (ms->possible_cpus) {
2884         assert(ms->possible_cpus->len == max_cpus);
2885         return ms->possible_cpus;
2886     }
2887 
2888     ms->possible_cpus = g_malloc0(sizeof(CPUArchIdList) +
2889                                   sizeof(CPUArchId) * max_cpus);
2890     ms->possible_cpus->len = max_cpus;
2891     for (n = 0; n < ms->possible_cpus->len; n++) {
2892         ms->possible_cpus->cpus[n].type = ms->cpu_type;
2893         ms->possible_cpus->cpus[n].arch_id =
2894             virt_cpu_mp_affinity(vms, n);
2895 
2896         assert(!mc->smp_props.dies_supported);
2897         ms->possible_cpus->cpus[n].props.has_socket_id = true;
2898         ms->possible_cpus->cpus[n].props.socket_id =
2899             n / (ms->smp.clusters * ms->smp.cores * ms->smp.threads);
2900         ms->possible_cpus->cpus[n].props.has_cluster_id = true;
2901         ms->possible_cpus->cpus[n].props.cluster_id =
2902             (n / (ms->smp.cores * ms->smp.threads)) % ms->smp.clusters;
2903         ms->possible_cpus->cpus[n].props.has_core_id = true;
2904         ms->possible_cpus->cpus[n].props.core_id =
2905             (n / ms->smp.threads) % ms->smp.cores;
2906         ms->possible_cpus->cpus[n].props.has_thread_id = true;
2907         ms->possible_cpus->cpus[n].props.thread_id =
2908             n % ms->smp.threads;
2909     }
2910     return ms->possible_cpus;
2911 }
2912 
virt_memory_pre_plug(HotplugHandler * hotplug_dev,DeviceState * dev,Error ** errp)2913 static void virt_memory_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
2914                                  Error **errp)
2915 {
2916     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2917     const MachineState *ms = MACHINE(hotplug_dev);
2918     const bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2919 
2920     if (!vms->acpi_dev) {
2921         error_setg(errp,
2922                    "memory hotplug is not enabled: missing acpi-ged device");
2923         return;
2924     }
2925 
2926     if (vms->mte) {
2927         error_setg(errp, "memory hotplug is not enabled: MTE is enabled");
2928         return;
2929     }
2930 
2931     if (is_nvdimm && !ms->nvdimms_state->is_enabled) {
2932         error_setg(errp, "nvdimm is not enabled: add 'nvdimm=on' to '-M'");
2933         return;
2934     }
2935 
2936     pc_dimm_pre_plug(PC_DIMM(dev), MACHINE(hotplug_dev), errp);
2937 }
2938 
virt_memory_plug(HotplugHandler * hotplug_dev,DeviceState * dev,Error ** errp)2939 static void virt_memory_plug(HotplugHandler *hotplug_dev,
2940                              DeviceState *dev, Error **errp)
2941 {
2942     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2943     MachineState *ms = MACHINE(hotplug_dev);
2944     bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2945 
2946     pc_dimm_plug(PC_DIMM(dev), MACHINE(vms));
2947 
2948     if (is_nvdimm) {
2949         nvdimm_plug(ms->nvdimms_state);
2950     }
2951 
2952     hotplug_handler_plug(HOTPLUG_HANDLER(vms->acpi_dev),
2953                          dev, &error_abort);
2954 }
2955 
virt_machine_device_pre_plug_cb(HotplugHandler * hotplug_dev,DeviceState * dev,Error ** errp)2956 static void virt_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev,
2957                                             DeviceState *dev, Error **errp)
2958 {
2959     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2960 
2961     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2962         virt_memory_pre_plug(hotplug_dev, dev, errp);
2963     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI)) {
2964         virtio_md_pci_pre_plug(VIRTIO_MD_PCI(dev), MACHINE(hotplug_dev), errp);
2965     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2966         hwaddr db_start = 0, db_end = 0;
2967         QList *reserved_regions;
2968         char *resv_prop_str;
2969 
2970         if (vms->iommu != VIRT_IOMMU_NONE) {
2971             error_setg(errp, "virt machine does not support multiple IOMMUs");
2972             return;
2973         }
2974 
2975         switch (vms->msi_controller) {
2976         case VIRT_MSI_CTRL_NONE:
2977             return;
2978         case VIRT_MSI_CTRL_ITS:
2979             /* GITS_TRANSLATER page */
2980             db_start = base_memmap[VIRT_GIC_ITS].base + 0x10000;
2981             db_end = base_memmap[VIRT_GIC_ITS].base +
2982                      base_memmap[VIRT_GIC_ITS].size - 1;
2983             break;
2984         case VIRT_MSI_CTRL_GICV2M:
2985             /* MSI_SETSPI_NS page */
2986             db_start = base_memmap[VIRT_GIC_V2M].base;
2987             db_end = db_start + base_memmap[VIRT_GIC_V2M].size - 1;
2988             break;
2989         }
2990         resv_prop_str = g_strdup_printf("0x%"PRIx64":0x%"PRIx64":%u",
2991                                         db_start, db_end,
2992                                         VIRTIO_IOMMU_RESV_MEM_T_MSI);
2993 
2994         reserved_regions = qlist_new();
2995         qlist_append_str(reserved_regions, resv_prop_str);
2996         qdev_prop_set_array(dev, "reserved-regions", reserved_regions);
2997         g_free(resv_prop_str);
2998     }
2999 }
3000 
virt_machine_device_plug_cb(HotplugHandler * hotplug_dev,DeviceState * dev,Error ** errp)3001 static void virt_machine_device_plug_cb(HotplugHandler *hotplug_dev,
3002                                         DeviceState *dev, Error **errp)
3003 {
3004     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
3005 
3006     if (vms->platform_bus_dev) {
3007         MachineClass *mc = MACHINE_GET_CLASS(vms);
3008 
3009         if (device_is_dynamic_sysbus(mc, dev)) {
3010             platform_bus_link_device(PLATFORM_BUS_DEVICE(vms->platform_bus_dev),
3011                                      SYS_BUS_DEVICE(dev));
3012         }
3013     }
3014 
3015     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
3016         virt_memory_plug(hotplug_dev, dev, errp);
3017     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI)) {
3018         virtio_md_pci_plug(VIRTIO_MD_PCI(dev), MACHINE(hotplug_dev), errp);
3019     }
3020 
3021     if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
3022         PCIDevice *pdev = PCI_DEVICE(dev);
3023 
3024         vms->iommu = VIRT_IOMMU_VIRTIO;
3025         vms->virtio_iommu_bdf = pci_get_bdf(pdev);
3026         create_virtio_iommu_dt_bindings(vms);
3027     }
3028 }
3029 
virt_dimm_unplug_request(HotplugHandler * hotplug_dev,DeviceState * dev,Error ** errp)3030 static void virt_dimm_unplug_request(HotplugHandler *hotplug_dev,
3031                                      DeviceState *dev, Error **errp)
3032 {
3033     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
3034 
3035     if (!vms->acpi_dev) {
3036         error_setg(errp,
3037                    "memory hotplug is not enabled: missing acpi-ged device");
3038         return;
3039     }
3040 
3041     if (object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM)) {
3042         error_setg(errp, "nvdimm device hot unplug is not supported yet.");
3043         return;
3044     }
3045 
3046     hotplug_handler_unplug_request(HOTPLUG_HANDLER(vms->acpi_dev), dev,
3047                                    errp);
3048 }
3049 
virt_dimm_unplug(HotplugHandler * hotplug_dev,DeviceState * dev,Error ** errp)3050 static void virt_dimm_unplug(HotplugHandler *hotplug_dev,
3051                              DeviceState *dev, Error **errp)
3052 {
3053     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
3054     Error *local_err = NULL;
3055 
3056     hotplug_handler_unplug(HOTPLUG_HANDLER(vms->acpi_dev), dev, &local_err);
3057     if (local_err) {
3058         goto out;
3059     }
3060 
3061     pc_dimm_unplug(PC_DIMM(dev), MACHINE(vms));
3062     qdev_unrealize(dev);
3063 
3064 out:
3065     error_propagate(errp, local_err);
3066 }
3067 
virt_machine_device_unplug_request_cb(HotplugHandler * hotplug_dev,DeviceState * dev,Error ** errp)3068 static void virt_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
3069                                           DeviceState *dev, Error **errp)
3070 {
3071     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
3072         virt_dimm_unplug_request(hotplug_dev, dev, errp);
3073     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI)) {
3074         virtio_md_pci_unplug_request(VIRTIO_MD_PCI(dev), MACHINE(hotplug_dev),
3075                                      errp);
3076     } else {
3077         error_setg(errp, "device unplug request for unsupported device"
3078                    " type: %s", object_get_typename(OBJECT(dev)));
3079     }
3080 }
3081 
virt_machine_device_unplug_cb(HotplugHandler * hotplug_dev,DeviceState * dev,Error ** errp)3082 static void virt_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
3083                                           DeviceState *dev, Error **errp)
3084 {
3085     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
3086         virt_dimm_unplug(hotplug_dev, dev, errp);
3087     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI)) {
3088         virtio_md_pci_unplug(VIRTIO_MD_PCI(dev), MACHINE(hotplug_dev), errp);
3089     } else {
3090         error_setg(errp, "virt: device unplug for unsupported device"
3091                    " type: %s", object_get_typename(OBJECT(dev)));
3092     }
3093 }
3094 
virt_machine_get_hotplug_handler(MachineState * machine,DeviceState * dev)3095 static HotplugHandler *virt_machine_get_hotplug_handler(MachineState *machine,
3096                                                         DeviceState *dev)
3097 {
3098     MachineClass *mc = MACHINE_GET_CLASS(machine);
3099 
3100     if (device_is_dynamic_sysbus(mc, dev) ||
3101         object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) ||
3102         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI) ||
3103         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
3104         return HOTPLUG_HANDLER(machine);
3105     }
3106     return NULL;
3107 }
3108 
3109 /*
3110  * for arm64 kvm_type [7-0] encodes the requested number of bits
3111  * in the IPA address space
3112  */
virt_kvm_type(MachineState * ms,const char * type_str)3113 static int virt_kvm_type(MachineState *ms, const char *type_str)
3114 {
3115     VirtMachineState *vms = VIRT_MACHINE(ms);
3116     int max_vm_pa_size, requested_pa_size;
3117     bool fixed_ipa;
3118 
3119     max_vm_pa_size = kvm_arm_get_max_vm_ipa_size(ms, &fixed_ipa);
3120 
3121     /* we freeze the memory map to compute the highest gpa */
3122     virt_set_memmap(vms, max_vm_pa_size);
3123 
3124     requested_pa_size = 64 - clz64(vms->highest_gpa);
3125 
3126     /*
3127      * KVM requires the IPA size to be at least 32 bits.
3128      */
3129     if (requested_pa_size < 32) {
3130         requested_pa_size = 32;
3131     }
3132 
3133     if (requested_pa_size > max_vm_pa_size) {
3134         error_report("-m and ,maxmem option values "
3135                      "require an IPA range (%d bits) larger than "
3136                      "the one supported by the host (%d bits)",
3137                      requested_pa_size, max_vm_pa_size);
3138         return -1;
3139     }
3140     /*
3141      * We return the requested PA log size, unless KVM only supports
3142      * the implicit legacy 40b IPA setting, in which case the kvm_type
3143      * must be 0.
3144      */
3145     return fixed_ipa ? 0 : requested_pa_size;
3146 }
3147 
virt_hvf_get_physical_address_range(MachineState * ms)3148 static int virt_hvf_get_physical_address_range(MachineState *ms)
3149 {
3150     VirtMachineState *vms = VIRT_MACHINE(ms);
3151 
3152     int default_ipa_size = hvf_arm_get_default_ipa_bit_size();
3153     int max_ipa_size = hvf_arm_get_max_ipa_bit_size();
3154 
3155     /* We freeze the memory map to compute the highest gpa */
3156     virt_set_memmap(vms, max_ipa_size);
3157 
3158     int requested_ipa_size = 64 - clz64(vms->highest_gpa);
3159 
3160     /*
3161      * If we're <= the default IPA size just use the default.
3162      * If we're above the default but below the maximum, round up to
3163      * the maximum. hvf_arm_get_max_ipa_bit_size() conveniently only
3164      * returns values that are valid ARM PARange values.
3165      */
3166     if (requested_ipa_size <= default_ipa_size) {
3167         requested_ipa_size = default_ipa_size;
3168     } else if (requested_ipa_size <= max_ipa_size) {
3169         requested_ipa_size = max_ipa_size;
3170     } else {
3171         error_report("-m and ,maxmem option values "
3172                      "require an IPA range (%d bits) larger than "
3173                      "the one supported by the host (%d bits)",
3174                      requested_ipa_size, max_ipa_size);
3175         return -1;
3176     }
3177 
3178     return requested_ipa_size;
3179 }
3180 
virt_machine_class_init(ObjectClass * oc,const void * data)3181 static void virt_machine_class_init(ObjectClass *oc, const void *data)
3182 {
3183     MachineClass *mc = MACHINE_CLASS(oc);
3184     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
3185     static const char * const valid_cpu_types[] = {
3186 #ifdef CONFIG_TCG
3187         ARM_CPU_TYPE_NAME("cortex-a7"),
3188         ARM_CPU_TYPE_NAME("cortex-a15"),
3189 #ifdef TARGET_AARCH64
3190         ARM_CPU_TYPE_NAME("cortex-a35"),
3191         ARM_CPU_TYPE_NAME("cortex-a55"),
3192         ARM_CPU_TYPE_NAME("cortex-a72"),
3193         ARM_CPU_TYPE_NAME("cortex-a76"),
3194         ARM_CPU_TYPE_NAME("cortex-a710"),
3195         ARM_CPU_TYPE_NAME("a64fx"),
3196         ARM_CPU_TYPE_NAME("neoverse-n1"),
3197         ARM_CPU_TYPE_NAME("neoverse-v1"),
3198         ARM_CPU_TYPE_NAME("neoverse-n2"),
3199 #endif /* TARGET_AARCH64 */
3200 #endif /* CONFIG_TCG */
3201 #ifdef TARGET_AARCH64
3202         ARM_CPU_TYPE_NAME("cortex-a53"),
3203         ARM_CPU_TYPE_NAME("cortex-a57"),
3204 #if defined(CONFIG_KVM) || defined(CONFIG_HVF)
3205         ARM_CPU_TYPE_NAME("host"),
3206 #endif /* CONFIG_KVM || CONFIG_HVF */
3207 #endif /* TARGET_AARCH64 */
3208         ARM_CPU_TYPE_NAME("max"),
3209         NULL
3210     };
3211 
3212     mc->init = machvirt_init;
3213     /* Start with max_cpus set to 512, which is the maximum supported by KVM.
3214      * The value may be reduced later when we have more information about the
3215      * configuration of the particular instance.
3216      */
3217     mc->max_cpus = 512;
3218     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_CALXEDA_XGMAC);
3219     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_AMD_XGBE);
3220     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_RAMFB_DEVICE);
3221     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_PLATFORM);
3222     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_UEFI_VARS_SYSBUS);
3223 #ifdef CONFIG_TPM
3224     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_TPM_TIS_SYSBUS);
3225 #endif
3226     mc->block_default_type = IF_VIRTIO;
3227     mc->no_cdrom = 1;
3228     mc->pci_allow_0_address = true;
3229     /* We know we will never create a pre-ARMv7 CPU which needs 1K pages */
3230     mc->minimum_page_bits = 12;
3231     mc->possible_cpu_arch_ids = virt_possible_cpu_arch_ids;
3232     mc->cpu_index_to_instance_props = virt_cpu_index_to_props;
3233 #ifdef CONFIG_TCG
3234     mc->default_cpu_type = ARM_CPU_TYPE_NAME("cortex-a15");
3235 #else
3236     mc->default_cpu_type = ARM_CPU_TYPE_NAME("max");
3237 #endif
3238     mc->valid_cpu_types = valid_cpu_types;
3239     mc->get_default_cpu_node_id = virt_get_default_cpu_node_id;
3240     mc->kvm_type = virt_kvm_type;
3241     mc->hvf_get_physical_address_range = virt_hvf_get_physical_address_range;
3242     assert(!mc->get_hotplug_handler);
3243     mc->get_hotplug_handler = virt_machine_get_hotplug_handler;
3244     hc->pre_plug = virt_machine_device_pre_plug_cb;
3245     hc->plug = virt_machine_device_plug_cb;
3246     hc->unplug_request = virt_machine_device_unplug_request_cb;
3247     hc->unplug = virt_machine_device_unplug_cb;
3248     mc->nvdimm_supported = true;
3249     mc->smp_props.clusters_supported = true;
3250     mc->auto_enable_numa_with_memhp = true;
3251     mc->auto_enable_numa_with_memdev = true;
3252     /* platform instead of architectural choice */
3253     mc->cpu_cluster_has_numa_boundary = true;
3254     mc->default_ram_id = "mach-virt.ram";
3255     mc->default_nic = "virtio-net-pci";
3256 
3257     object_class_property_add(oc, "acpi", "OnOffAuto",
3258         virt_get_acpi, virt_set_acpi,
3259         NULL, NULL);
3260     object_class_property_set_description(oc, "acpi",
3261         "Enable ACPI");
3262     object_class_property_add_bool(oc, "secure", virt_get_secure,
3263                                    virt_set_secure);
3264     object_class_property_set_description(oc, "secure",
3265                                                 "Set on/off to enable/disable the ARM "
3266                                                 "Security Extensions (TrustZone)");
3267 
3268     object_class_property_add_bool(oc, "virtualization", virt_get_virt,
3269                                    virt_set_virt);
3270     object_class_property_set_description(oc, "virtualization",
3271                                           "Set on/off to enable/disable emulating a "
3272                                           "guest CPU which implements the ARM "
3273                                           "Virtualization Extensions");
3274 
3275     object_class_property_add_bool(oc, "highmem", virt_get_highmem,
3276                                    virt_set_highmem);
3277     object_class_property_set_description(oc, "highmem",
3278                                           "Set on/off to enable/disable using "
3279                                           "physical address space above 32 bits");
3280 
3281     object_class_property_add_bool(oc, "compact-highmem",
3282                                    virt_get_compact_highmem,
3283                                    virt_set_compact_highmem);
3284     object_class_property_set_description(oc, "compact-highmem",
3285                                           "Set on/off to enable/disable compact "
3286                                           "layout for high memory regions");
3287 
3288     object_class_property_add_bool(oc, "highmem-redists",
3289                                    virt_get_highmem_redists,
3290                                    virt_set_highmem_redists);
3291     object_class_property_set_description(oc, "highmem-redists",
3292                                           "Set on/off to enable/disable high "
3293                                           "memory region for GICv3 or GICv4 "
3294                                           "redistributor");
3295 
3296     object_class_property_add_bool(oc, "highmem-ecam",
3297                                    virt_get_highmem_ecam,
3298                                    virt_set_highmem_ecam);
3299     object_class_property_set_description(oc, "highmem-ecam",
3300                                           "Set on/off to enable/disable high "
3301                                           "memory region for PCI ECAM");
3302 
3303     object_class_property_add_bool(oc, "highmem-mmio",
3304                                    virt_get_highmem_mmio,
3305                                    virt_set_highmem_mmio);
3306     object_class_property_set_description(oc, "highmem-mmio",
3307                                           "Set on/off to enable/disable high "
3308                                           "memory region for PCI MMIO");
3309 
3310     object_class_property_add(oc, "highmem-mmio-size", "size",
3311                                    virt_get_highmem_mmio_size,
3312                                    virt_set_highmem_mmio_size,
3313                                    NULL, NULL);
3314     object_class_property_set_description(oc, "highmem-mmio-size",
3315                                           "Set the high memory region size "
3316                                           "for PCI MMIO");
3317 
3318     object_class_property_add_str(oc, "gic-version", virt_get_gic_version,
3319                                   virt_set_gic_version);
3320     object_class_property_set_description(oc, "gic-version",
3321                                           "Set GIC version. "
3322                                           "Valid values are 2, 3, 4, host and max");
3323 
3324     object_class_property_add_str(oc, "iommu", virt_get_iommu, virt_set_iommu);
3325     object_class_property_set_description(oc, "iommu",
3326                                           "Set the IOMMU type. "
3327                                           "Valid values are none and smmuv3");
3328 
3329     object_class_property_add_bool(oc, "default-bus-bypass-iommu",
3330                                    virt_get_default_bus_bypass_iommu,
3331                                    virt_set_default_bus_bypass_iommu);
3332     object_class_property_set_description(oc, "default-bus-bypass-iommu",
3333                                           "Set on/off to enable/disable "
3334                                           "bypass_iommu for default root bus");
3335 
3336     object_class_property_add_bool(oc, "ras", virt_get_ras,
3337                                    virt_set_ras);
3338     object_class_property_set_description(oc, "ras",
3339                                           "Set on/off to enable/disable reporting host memory errors "
3340                                           "to a KVM guest using ACPI and guest external abort exceptions");
3341 
3342     object_class_property_add_bool(oc, "mte", virt_get_mte, virt_set_mte);
3343     object_class_property_set_description(oc, "mte",
3344                                           "Set on/off to enable/disable emulating a "
3345                                           "guest CPU which implements the ARM "
3346                                           "Memory Tagging Extension");
3347 
3348     object_class_property_add_bool(oc, "its", virt_get_its,
3349                                    virt_set_its);
3350     object_class_property_set_description(oc, "its",
3351                                           "Set on/off to enable/disable "
3352                                           "ITS instantiation");
3353 
3354     object_class_property_add_bool(oc, "dtb-randomness",
3355                                    virt_get_dtb_randomness,
3356                                    virt_set_dtb_randomness);
3357     object_class_property_set_description(oc, "dtb-randomness",
3358                                           "Set off to disable passing random or "
3359                                           "non-deterministic dtb nodes to guest");
3360 
3361     object_class_property_add_bool(oc, "dtb-kaslr-seed",
3362                                    virt_get_dtb_randomness,
3363                                    virt_set_dtb_randomness);
3364     object_class_property_set_description(oc, "dtb-kaslr-seed",
3365                                           "Deprecated synonym of dtb-randomness");
3366 
3367     object_class_property_add_str(oc, "x-oem-id",
3368                                   virt_get_oem_id,
3369                                   virt_set_oem_id);
3370     object_class_property_set_description(oc, "x-oem-id",
3371                                           "Override the default value of field OEMID "
3372                                           "in ACPI table header."
3373                                           "The string may be up to 6 bytes in size");
3374 
3375 
3376     object_class_property_add_str(oc, "x-oem-table-id",
3377                                   virt_get_oem_table_id,
3378                                   virt_set_oem_table_id);
3379     object_class_property_set_description(oc, "x-oem-table-id",
3380                                           "Override the default value of field OEM Table ID "
3381                                           "in ACPI table header."
3382                                           "The string may be up to 8 bytes in size");
3383 
3384 }
3385 
virt_instance_init(Object * obj)3386 static void virt_instance_init(Object *obj)
3387 {
3388     VirtMachineState *vms = VIRT_MACHINE(obj);
3389     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
3390 
3391     /* EL3 is disabled by default on virt: this makes us consistent
3392      * between KVM and TCG for this board, and it also allows us to
3393      * boot UEFI blobs which assume no TrustZone support.
3394      */
3395     vms->secure = false;
3396 
3397     /* EL2 is also disabled by default, for similar reasons */
3398     vms->virt = false;
3399 
3400     /* High memory is enabled by default */
3401     vms->highmem = true;
3402     vms->highmem_compact = !vmc->no_highmem_compact;
3403     vms->gic_version = VIRT_GIC_VERSION_NOSEL;
3404 
3405     vms->highmem_ecam = true;
3406     vms->highmem_mmio = true;
3407     vms->highmem_redists = true;
3408 
3409     /* Default allows ITS instantiation */
3410     vms->its = true;
3411     /* Allow ITS emulation if the machine version supports it */
3412     vms->tcg_its = !vmc->no_tcg_its;
3413 
3414     /* Default disallows iommu instantiation */
3415     vms->iommu = VIRT_IOMMU_NONE;
3416 
3417     /* The default root bus is attached to iommu by default */
3418     vms->default_bus_bypass_iommu = false;
3419 
3420     /* Default disallows RAS instantiation */
3421     vms->ras = false;
3422 
3423     /* MTE is disabled by default.  */
3424     vms->mte = false;
3425 
3426     /* Supply kaslr-seed and rng-seed by default */
3427     vms->dtb_randomness = true;
3428 
3429     vms->irqmap = a15irqmap;
3430 
3431     virt_flash_create(vms);
3432 
3433     vms->oem_id = g_strndup(ACPI_BUILD_APPNAME6, 6);
3434     vms->oem_table_id = g_strndup(ACPI_BUILD_APPNAME8, 8);
3435     cxl_machine_init(obj, &vms->cxl_devices_state);
3436 }
3437 
3438 static const TypeInfo virt_machine_info = {
3439     .name          = TYPE_VIRT_MACHINE,
3440     .parent        = TYPE_MACHINE,
3441     .abstract      = true,
3442     .instance_size = sizeof(VirtMachineState),
3443     .class_size    = sizeof(VirtMachineClass),
3444     .class_init    = virt_machine_class_init,
3445     .instance_init = virt_instance_init,
3446     .interfaces = (const InterfaceInfo[]) {
3447          { TYPE_HOTPLUG_HANDLER },
3448          { }
3449     },
3450 };
3451 
machvirt_machine_init(void)3452 static void machvirt_machine_init(void)
3453 {
3454     type_register_static(&virt_machine_info);
3455 }
3456 type_init(machvirt_machine_init);
3457 
virt_machine_10_1_options(MachineClass * mc)3458 static void virt_machine_10_1_options(MachineClass *mc)
3459 {
3460 }
3461 DEFINE_VIRT_MACHINE_AS_LATEST(10, 1)
3462 
virt_machine_10_0_options(MachineClass * mc)3463 static void virt_machine_10_0_options(MachineClass *mc)
3464 {
3465     virt_machine_10_1_options(mc);
3466     compat_props_add(mc->compat_props, hw_compat_10_0, hw_compat_10_0_len);
3467 }
3468 DEFINE_VIRT_MACHINE(10, 0)
3469 
virt_machine_9_2_options(MachineClass * mc)3470 static void virt_machine_9_2_options(MachineClass *mc)
3471 {
3472     virt_machine_10_0_options(mc);
3473     compat_props_add(mc->compat_props, hw_compat_9_2, hw_compat_9_2_len);
3474 }
3475 DEFINE_VIRT_MACHINE(9, 2)
3476 
virt_machine_9_1_options(MachineClass * mc)3477 static void virt_machine_9_1_options(MachineClass *mc)
3478 {
3479     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3480 
3481     virt_machine_9_2_options(mc);
3482     compat_props_add(mc->compat_props, hw_compat_9_1, hw_compat_9_1_len);
3483     /* 9.1 and earlier have only a stage-1 SMMU, not a nested s1+2 one */
3484     vmc->no_nested_smmu = true;
3485 }
3486 DEFINE_VIRT_MACHINE(9, 1)
3487 
virt_machine_9_0_options(MachineClass * mc)3488 static void virt_machine_9_0_options(MachineClass *mc)
3489 {
3490     virt_machine_9_1_options(mc);
3491     mc->smbios_memory_device_size = 16 * GiB;
3492     compat_props_add(mc->compat_props, hw_compat_9_0, hw_compat_9_0_len);
3493 }
3494 DEFINE_VIRT_MACHINE(9, 0)
3495 
virt_machine_8_2_options(MachineClass * mc)3496 static void virt_machine_8_2_options(MachineClass *mc)
3497 {
3498     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3499 
3500     virt_machine_9_0_options(mc);
3501     compat_props_add(mc->compat_props, hw_compat_8_2, hw_compat_8_2_len);
3502     /*
3503      * Don't expose NS_EL2_VIRT timer IRQ in DTB on ACPI on 8.2 and
3504      * earlier machines. (Exposing it tickles a bug in older EDK2
3505      * guest BIOS binaries.)
3506      */
3507     vmc->no_ns_el2_virt_timer_irq = true;
3508 }
3509 DEFINE_VIRT_MACHINE(8, 2)
3510 
virt_machine_8_1_options(MachineClass * mc)3511 static void virt_machine_8_1_options(MachineClass *mc)
3512 {
3513     virt_machine_8_2_options(mc);
3514     compat_props_add(mc->compat_props, hw_compat_8_1, hw_compat_8_1_len);
3515 }
3516 DEFINE_VIRT_MACHINE(8, 1)
3517 
virt_machine_8_0_options(MachineClass * mc)3518 static void virt_machine_8_0_options(MachineClass *mc)
3519 {
3520     virt_machine_8_1_options(mc);
3521     compat_props_add(mc->compat_props, hw_compat_8_0, hw_compat_8_0_len);
3522 }
3523 DEFINE_VIRT_MACHINE(8, 0)
3524 
virt_machine_7_2_options(MachineClass * mc)3525 static void virt_machine_7_2_options(MachineClass *mc)
3526 {
3527     virt_machine_8_0_options(mc);
3528     compat_props_add(mc->compat_props, hw_compat_7_2, hw_compat_7_2_len);
3529 }
3530 DEFINE_VIRT_MACHINE(7, 2)
3531 
virt_machine_7_1_options(MachineClass * mc)3532 static void virt_machine_7_1_options(MachineClass *mc)
3533 {
3534     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3535 
3536     virt_machine_7_2_options(mc);
3537     compat_props_add(mc->compat_props, hw_compat_7_1, hw_compat_7_1_len);
3538     /* Compact layout for high memory regions was introduced with 7.2 */
3539     vmc->no_highmem_compact = true;
3540 }
3541 DEFINE_VIRT_MACHINE(7, 1)
3542 
virt_machine_7_0_options(MachineClass * mc)3543 static void virt_machine_7_0_options(MachineClass *mc)
3544 {
3545     virt_machine_7_1_options(mc);
3546     compat_props_add(mc->compat_props, hw_compat_7_0, hw_compat_7_0_len);
3547 }
3548 DEFINE_VIRT_MACHINE(7, 0)
3549 
virt_machine_6_2_options(MachineClass * mc)3550 static void virt_machine_6_2_options(MachineClass *mc)
3551 {
3552     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3553 
3554     virt_machine_7_0_options(mc);
3555     compat_props_add(mc->compat_props, hw_compat_6_2, hw_compat_6_2_len);
3556     vmc->no_tcg_lpa2 = true;
3557 }
3558 DEFINE_VIRT_MACHINE(6, 2)
3559 
virt_machine_6_1_options(MachineClass * mc)3560 static void virt_machine_6_1_options(MachineClass *mc)
3561 {
3562     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3563 
3564     virt_machine_6_2_options(mc);
3565     compat_props_add(mc->compat_props, hw_compat_6_1, hw_compat_6_1_len);
3566     mc->smp_props.prefer_sockets = true;
3567     vmc->no_cpu_topology = true;
3568 
3569     /* qemu ITS was introduced with 6.2 */
3570     vmc->no_tcg_its = true;
3571 }
3572 DEFINE_VIRT_MACHINE(6, 1)
3573 
virt_machine_6_0_options(MachineClass * mc)3574 static void virt_machine_6_0_options(MachineClass *mc)
3575 {
3576     virt_machine_6_1_options(mc);
3577     compat_props_add(mc->compat_props, hw_compat_6_0, hw_compat_6_0_len);
3578 }
3579 DEFINE_VIRT_MACHINE(6, 0)
3580 
virt_machine_5_2_options(MachineClass * mc)3581 static void virt_machine_5_2_options(MachineClass *mc)
3582 {
3583     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3584 
3585     virt_machine_6_0_options(mc);
3586     compat_props_add(mc->compat_props, hw_compat_5_2, hw_compat_5_2_len);
3587     vmc->no_secure_gpio = true;
3588 }
3589 DEFINE_VIRT_MACHINE(5, 2)
3590 
virt_machine_5_1_options(MachineClass * mc)3591 static void virt_machine_5_1_options(MachineClass *mc)
3592 {
3593     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3594 
3595     virt_machine_5_2_options(mc);
3596     compat_props_add(mc->compat_props, hw_compat_5_1, hw_compat_5_1_len);
3597     vmc->no_kvm_steal_time = true;
3598 }
3599 DEFINE_VIRT_MACHINE(5, 1)
3600 
virt_machine_5_0_options(MachineClass * mc)3601 static void virt_machine_5_0_options(MachineClass *mc)
3602 {
3603     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3604 
3605     virt_machine_5_1_options(mc);
3606     compat_props_add(mc->compat_props, hw_compat_5_0, hw_compat_5_0_len);
3607     mc->numa_mem_supported = true;
3608     vmc->acpi_expose_flash = true;
3609     mc->auto_enable_numa_with_memdev = false;
3610 }
3611 DEFINE_VIRT_MACHINE(5, 0)
3612 
virt_machine_4_2_options(MachineClass * mc)3613 static void virt_machine_4_2_options(MachineClass *mc)
3614 {
3615     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3616 
3617     virt_machine_5_0_options(mc);
3618     compat_props_add(mc->compat_props, hw_compat_4_2, hw_compat_4_2_len);
3619     vmc->kvm_no_adjvtime = true;
3620 }
3621 DEFINE_VIRT_MACHINE(4, 2)
3622 
virt_machine_4_1_options(MachineClass * mc)3623 static void virt_machine_4_1_options(MachineClass *mc)
3624 {
3625     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3626 
3627     virt_machine_4_2_options(mc);
3628     compat_props_add(mc->compat_props, hw_compat_4_1, hw_compat_4_1_len);
3629     vmc->no_ged = true;
3630     mc->auto_enable_numa_with_memhp = false;
3631 }
3632 DEFINE_VIRT_MACHINE(4, 1)
3633