xref: /openbmc/qemu/hw/ppc/spapr.c (revision 61b9251a)
1 /*
2  * QEMU PowerPC pSeries Logical Partition (aka sPAPR) hardware System Emulator
3  *
4  * Copyright (c) 2004-2007 Fabrice Bellard
5  * Copyright (c) 2007 Jocelyn Mayer
6  * Copyright (c) 2010 David Gibson, IBM Corporation.
7  *
8  * Permission is hereby granted, free of charge, to any person obtaining a copy
9  * of this software and associated documentation files (the "Software"), to deal
10  * in the Software without restriction, including without limitation the rights
11  * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
12  * copies of the Software, and to permit persons to whom the Software is
13  * furnished to do so, subject to the following conditions:
14  *
15  * The above copyright notice and this permission notice shall be included in
16  * all copies or substantial portions of the Software.
17  *
18  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
19  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
20  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
21  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
22  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
23  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
24  * THE SOFTWARE.
25  *
26  */
27 #include "sysemu/sysemu.h"
28 #include "sysemu/numa.h"
29 #include "hw/hw.h"
30 #include "hw/fw-path-provider.h"
31 #include "elf.h"
32 #include "net/net.h"
33 #include "sysemu/device_tree.h"
34 #include "sysemu/block-backend.h"
35 #include "sysemu/cpus.h"
36 #include "sysemu/kvm.h"
37 #include "sysemu/device_tree.h"
38 #include "kvm_ppc.h"
39 #include "migration/migration.h"
40 #include "mmu-hash64.h"
41 #include "qom/cpu.h"
42 
43 #include "hw/boards.h"
44 #include "hw/ppc/ppc.h"
45 #include "hw/loader.h"
46 
47 #include "hw/ppc/spapr.h"
48 #include "hw/ppc/spapr_vio.h"
49 #include "hw/pci-host/spapr.h"
50 #include "hw/ppc/xics.h"
51 #include "hw/pci/msi.h"
52 
53 #include "hw/pci/pci.h"
54 #include "hw/scsi/scsi.h"
55 #include "hw/virtio/virtio-scsi.h"
56 
57 #include "exec/address-spaces.h"
58 #include "hw/usb.h"
59 #include "qemu/config-file.h"
60 #include "qemu/error-report.h"
61 #include "trace.h"
62 #include "hw/nmi.h"
63 
64 #include "hw/compat.h"
65 #include "qemu-common.h"
66 
67 #include <libfdt.h>
68 
69 /* SLOF memory layout:
70  *
71  * SLOF raw image loaded at 0, copies its romfs right below the flat
72  * device-tree, then position SLOF itself 31M below that
73  *
74  * So we set FW_OVERHEAD to 40MB which should account for all of that
75  * and more
76  *
77  * We load our kernel at 4M, leaving space for SLOF initial image
78  */
79 #define FDT_MAX_SIZE            0x100000
80 #define RTAS_MAX_SIZE           0x10000
81 #define RTAS_MAX_ADDR           0x80000000 /* RTAS must stay below that */
82 #define FW_MAX_SIZE             0x400000
83 #define FW_FILE_NAME            "slof.bin"
84 #define FW_OVERHEAD             0x2800000
85 #define KERNEL_LOAD_ADDR        FW_MAX_SIZE
86 
87 #define MIN_RMA_SLOF            128UL
88 
89 #define TIMEBASE_FREQ           512000000ULL
90 
91 #define PHANDLE_XICP            0x00001111
92 
93 #define HTAB_SIZE(spapr)        (1ULL << ((spapr)->htab_shift))
94 
95 static XICSState *try_create_xics(const char *type, int nr_servers,
96                                   int nr_irqs, Error **errp)
97 {
98     Error *err = NULL;
99     DeviceState *dev;
100 
101     dev = qdev_create(NULL, type);
102     qdev_prop_set_uint32(dev, "nr_servers", nr_servers);
103     qdev_prop_set_uint32(dev, "nr_irqs", nr_irqs);
104     object_property_set_bool(OBJECT(dev), true, "realized", &err);
105     if (err) {
106         error_propagate(errp, err);
107         object_unparent(OBJECT(dev));
108         return NULL;
109     }
110     return XICS_COMMON(dev);
111 }
112 
113 static XICSState *xics_system_init(MachineState *machine,
114                                    int nr_servers, int nr_irqs)
115 {
116     XICSState *icp = NULL;
117 
118     if (kvm_enabled()) {
119         Error *err = NULL;
120 
121         if (machine_kernel_irqchip_allowed(machine)) {
122             icp = try_create_xics(TYPE_KVM_XICS, nr_servers, nr_irqs, &err);
123         }
124         if (machine_kernel_irqchip_required(machine) && !icp) {
125             error_report("kernel_irqchip requested but unavailable: %s",
126                          error_get_pretty(err));
127         }
128     }
129 
130     if (!icp) {
131         icp = try_create_xics(TYPE_XICS, nr_servers, nr_irqs, &error_abort);
132     }
133 
134     return icp;
135 }
136 
137 static int spapr_fixup_cpu_smt_dt(void *fdt, int offset, PowerPCCPU *cpu,
138                                   int smt_threads)
139 {
140     int i, ret = 0;
141     uint32_t servers_prop[smt_threads];
142     uint32_t gservers_prop[smt_threads * 2];
143     int index = ppc_get_vcpu_dt_id(cpu);
144 
145     if (cpu->cpu_version) {
146         ret = fdt_setprop_cell(fdt, offset, "cpu-version", cpu->cpu_version);
147         if (ret < 0) {
148             return ret;
149         }
150     }
151 
152     /* Build interrupt servers and gservers properties */
153     for (i = 0; i < smt_threads; i++) {
154         servers_prop[i] = cpu_to_be32(index + i);
155         /* Hack, direct the group queues back to cpu 0 */
156         gservers_prop[i*2] = cpu_to_be32(index + i);
157         gservers_prop[i*2 + 1] = 0;
158     }
159     ret = fdt_setprop(fdt, offset, "ibm,ppc-interrupt-server#s",
160                       servers_prop, sizeof(servers_prop));
161     if (ret < 0) {
162         return ret;
163     }
164     ret = fdt_setprop(fdt, offset, "ibm,ppc-interrupt-gserver#s",
165                       gservers_prop, sizeof(gservers_prop));
166 
167     return ret;
168 }
169 
170 static int spapr_fixup_cpu_numa_dt(void *fdt, int offset, CPUState *cs)
171 {
172     int ret = 0;
173     PowerPCCPU *cpu = POWERPC_CPU(cs);
174     int index = ppc_get_vcpu_dt_id(cpu);
175     uint32_t associativity[] = {cpu_to_be32(0x5),
176                                 cpu_to_be32(0x0),
177                                 cpu_to_be32(0x0),
178                                 cpu_to_be32(0x0),
179                                 cpu_to_be32(cs->numa_node),
180                                 cpu_to_be32(index)};
181 
182     /* Advertise NUMA via ibm,associativity */
183     if (nb_numa_nodes > 1) {
184         ret = fdt_setprop(fdt, offset, "ibm,associativity", associativity,
185                           sizeof(associativity));
186     }
187 
188     return ret;
189 }
190 
191 static int spapr_fixup_cpu_dt(void *fdt, sPAPRMachineState *spapr)
192 {
193     int ret = 0, offset, cpus_offset;
194     CPUState *cs;
195     char cpu_model[32];
196     int smt = kvmppc_smt_threads();
197     uint32_t pft_size_prop[] = {0, cpu_to_be32(spapr->htab_shift)};
198 
199     CPU_FOREACH(cs) {
200         PowerPCCPU *cpu = POWERPC_CPU(cs);
201         DeviceClass *dc = DEVICE_GET_CLASS(cs);
202         int index = ppc_get_vcpu_dt_id(cpu);
203 
204         if ((index % smt) != 0) {
205             continue;
206         }
207 
208         snprintf(cpu_model, 32, "%s@%x", dc->fw_name, index);
209 
210         cpus_offset = fdt_path_offset(fdt, "/cpus");
211         if (cpus_offset < 0) {
212             cpus_offset = fdt_add_subnode(fdt, fdt_path_offset(fdt, "/"),
213                                           "cpus");
214             if (cpus_offset < 0) {
215                 return cpus_offset;
216             }
217         }
218         offset = fdt_subnode_offset(fdt, cpus_offset, cpu_model);
219         if (offset < 0) {
220             offset = fdt_add_subnode(fdt, cpus_offset, cpu_model);
221             if (offset < 0) {
222                 return offset;
223             }
224         }
225 
226         ret = fdt_setprop(fdt, offset, "ibm,pft-size",
227                           pft_size_prop, sizeof(pft_size_prop));
228         if (ret < 0) {
229             return ret;
230         }
231 
232         ret = spapr_fixup_cpu_numa_dt(fdt, offset, cs);
233         if (ret < 0) {
234             return ret;
235         }
236 
237         ret = spapr_fixup_cpu_smt_dt(fdt, offset, cpu,
238                                      ppc_get_compat_smt_threads(cpu));
239         if (ret < 0) {
240             return ret;
241         }
242     }
243     return ret;
244 }
245 
246 
247 static size_t create_page_sizes_prop(CPUPPCState *env, uint32_t *prop,
248                                      size_t maxsize)
249 {
250     size_t maxcells = maxsize / sizeof(uint32_t);
251     int i, j, count;
252     uint32_t *p = prop;
253 
254     for (i = 0; i < PPC_PAGE_SIZES_MAX_SZ; i++) {
255         struct ppc_one_seg_page_size *sps = &env->sps.sps[i];
256 
257         if (!sps->page_shift) {
258             break;
259         }
260         for (count = 0; count < PPC_PAGE_SIZES_MAX_SZ; count++) {
261             if (sps->enc[count].page_shift == 0) {
262                 break;
263             }
264         }
265         if ((p - prop) >= (maxcells - 3 - count * 2)) {
266             break;
267         }
268         *(p++) = cpu_to_be32(sps->page_shift);
269         *(p++) = cpu_to_be32(sps->slb_enc);
270         *(p++) = cpu_to_be32(count);
271         for (j = 0; j < count; j++) {
272             *(p++) = cpu_to_be32(sps->enc[j].page_shift);
273             *(p++) = cpu_to_be32(sps->enc[j].pte_enc);
274         }
275     }
276 
277     return (p - prop) * sizeof(uint32_t);
278 }
279 
280 static hwaddr spapr_node0_size(void)
281 {
282     MachineState *machine = MACHINE(qdev_get_machine());
283 
284     if (nb_numa_nodes) {
285         int i;
286         for (i = 0; i < nb_numa_nodes; ++i) {
287             if (numa_info[i].node_mem) {
288                 return MIN(pow2floor(numa_info[i].node_mem),
289                            machine->ram_size);
290             }
291         }
292     }
293     return machine->ram_size;
294 }
295 
296 #define _FDT(exp) \
297     do { \
298         int ret = (exp);                                           \
299         if (ret < 0) {                                             \
300             fprintf(stderr, "qemu: error creating device tree: %s: %s\n", \
301                     #exp, fdt_strerror(ret));                      \
302             exit(1);                                               \
303         }                                                          \
304     } while (0)
305 
306 static void add_str(GString *s, const gchar *s1)
307 {
308     g_string_append_len(s, s1, strlen(s1) + 1);
309 }
310 
311 static void *spapr_create_fdt_skel(hwaddr initrd_base,
312                                    hwaddr initrd_size,
313                                    hwaddr kernel_size,
314                                    bool little_endian,
315                                    const char *kernel_cmdline,
316                                    uint32_t epow_irq)
317 {
318     void *fdt;
319     uint32_t start_prop = cpu_to_be32(initrd_base);
320     uint32_t end_prop = cpu_to_be32(initrd_base + initrd_size);
321     GString *hypertas = g_string_sized_new(256);
322     GString *qemu_hypertas = g_string_sized_new(256);
323     uint32_t refpoints[] = {cpu_to_be32(0x4), cpu_to_be32(0x4)};
324     uint32_t interrupt_server_ranges_prop[] = {0, cpu_to_be32(max_cpus)};
325     unsigned char vec5[] = {0x0, 0x0, 0x0, 0x0, 0x0, 0x80};
326     char *buf;
327 
328     add_str(hypertas, "hcall-pft");
329     add_str(hypertas, "hcall-term");
330     add_str(hypertas, "hcall-dabr");
331     add_str(hypertas, "hcall-interrupt");
332     add_str(hypertas, "hcall-tce");
333     add_str(hypertas, "hcall-vio");
334     add_str(hypertas, "hcall-splpar");
335     add_str(hypertas, "hcall-bulk");
336     add_str(hypertas, "hcall-set-mode");
337     add_str(qemu_hypertas, "hcall-memop1");
338 
339     fdt = g_malloc0(FDT_MAX_SIZE);
340     _FDT((fdt_create(fdt, FDT_MAX_SIZE)));
341 
342     if (kernel_size) {
343         _FDT((fdt_add_reservemap_entry(fdt, KERNEL_LOAD_ADDR, kernel_size)));
344     }
345     if (initrd_size) {
346         _FDT((fdt_add_reservemap_entry(fdt, initrd_base, initrd_size)));
347     }
348     _FDT((fdt_finish_reservemap(fdt)));
349 
350     /* Root node */
351     _FDT((fdt_begin_node(fdt, "")));
352     _FDT((fdt_property_string(fdt, "device_type", "chrp")));
353     _FDT((fdt_property_string(fdt, "model", "IBM pSeries (emulated by qemu)")));
354     _FDT((fdt_property_string(fdt, "compatible", "qemu,pseries")));
355 
356     /*
357      * Add info to guest to indentify which host is it being run on
358      * and what is the uuid of the guest
359      */
360     if (kvmppc_get_host_model(&buf)) {
361         _FDT((fdt_property_string(fdt, "host-model", buf)));
362         g_free(buf);
363     }
364     if (kvmppc_get_host_serial(&buf)) {
365         _FDT((fdt_property_string(fdt, "host-serial", buf)));
366         g_free(buf);
367     }
368 
369     buf = g_strdup_printf(UUID_FMT, qemu_uuid[0], qemu_uuid[1],
370                           qemu_uuid[2], qemu_uuid[3], qemu_uuid[4],
371                           qemu_uuid[5], qemu_uuid[6], qemu_uuid[7],
372                           qemu_uuid[8], qemu_uuid[9], qemu_uuid[10],
373                           qemu_uuid[11], qemu_uuid[12], qemu_uuid[13],
374                           qemu_uuid[14], qemu_uuid[15]);
375 
376     _FDT((fdt_property_string(fdt, "vm,uuid", buf)));
377     g_free(buf);
378 
379     if (qemu_get_vm_name()) {
380         _FDT((fdt_property_string(fdt, "ibm,partition-name",
381                                   qemu_get_vm_name())));
382     }
383 
384     _FDT((fdt_property_cell(fdt, "#address-cells", 0x2)));
385     _FDT((fdt_property_cell(fdt, "#size-cells", 0x2)));
386 
387     /* /chosen */
388     _FDT((fdt_begin_node(fdt, "chosen")));
389 
390     /* Set Form1_affinity */
391     _FDT((fdt_property(fdt, "ibm,architecture-vec-5", vec5, sizeof(vec5))));
392 
393     _FDT((fdt_property_string(fdt, "bootargs", kernel_cmdline)));
394     _FDT((fdt_property(fdt, "linux,initrd-start",
395                        &start_prop, sizeof(start_prop))));
396     _FDT((fdt_property(fdt, "linux,initrd-end",
397                        &end_prop, sizeof(end_prop))));
398     if (kernel_size) {
399         uint64_t kprop[2] = { cpu_to_be64(KERNEL_LOAD_ADDR),
400                               cpu_to_be64(kernel_size) };
401 
402         _FDT((fdt_property(fdt, "qemu,boot-kernel", &kprop, sizeof(kprop))));
403         if (little_endian) {
404             _FDT((fdt_property(fdt, "qemu,boot-kernel-le", NULL, 0)));
405         }
406     }
407     if (boot_menu) {
408         _FDT((fdt_property_cell(fdt, "qemu,boot-menu", boot_menu)));
409     }
410     _FDT((fdt_property_cell(fdt, "qemu,graphic-width", graphic_width)));
411     _FDT((fdt_property_cell(fdt, "qemu,graphic-height", graphic_height)));
412     _FDT((fdt_property_cell(fdt, "qemu,graphic-depth", graphic_depth)));
413 
414     _FDT((fdt_end_node(fdt)));
415 
416     /* RTAS */
417     _FDT((fdt_begin_node(fdt, "rtas")));
418 
419     if (!kvm_enabled() || kvmppc_spapr_use_multitce()) {
420         add_str(hypertas, "hcall-multi-tce");
421     }
422     _FDT((fdt_property(fdt, "ibm,hypertas-functions", hypertas->str,
423                        hypertas->len)));
424     g_string_free(hypertas, TRUE);
425     _FDT((fdt_property(fdt, "qemu,hypertas-functions", qemu_hypertas->str,
426                        qemu_hypertas->len)));
427     g_string_free(qemu_hypertas, TRUE);
428 
429     _FDT((fdt_property(fdt, "ibm,associativity-reference-points",
430         refpoints, sizeof(refpoints))));
431 
432     _FDT((fdt_property_cell(fdt, "rtas-error-log-max", RTAS_ERROR_LOG_MAX)));
433     _FDT((fdt_property_cell(fdt, "rtas-event-scan-rate",
434                             RTAS_EVENT_SCAN_RATE)));
435 
436     if (msi_supported) {
437         _FDT((fdt_property(fdt, "ibm,change-msix-capable", NULL, 0)));
438     }
439 
440     /*
441      * According to PAPR, rtas ibm,os-term does not guarantee a return
442      * back to the guest cpu.
443      *
444      * While an additional ibm,extended-os-term property indicates that
445      * rtas call return will always occur. Set this property.
446      */
447     _FDT((fdt_property(fdt, "ibm,extended-os-term", NULL, 0)));
448 
449     _FDT((fdt_end_node(fdt)));
450 
451     /* interrupt controller */
452     _FDT((fdt_begin_node(fdt, "interrupt-controller")));
453 
454     _FDT((fdt_property_string(fdt, "device_type",
455                               "PowerPC-External-Interrupt-Presentation")));
456     _FDT((fdt_property_string(fdt, "compatible", "IBM,ppc-xicp")));
457     _FDT((fdt_property(fdt, "interrupt-controller", NULL, 0)));
458     _FDT((fdt_property(fdt, "ibm,interrupt-server-ranges",
459                        interrupt_server_ranges_prop,
460                        sizeof(interrupt_server_ranges_prop))));
461     _FDT((fdt_property_cell(fdt, "#interrupt-cells", 2)));
462     _FDT((fdt_property_cell(fdt, "linux,phandle", PHANDLE_XICP)));
463     _FDT((fdt_property_cell(fdt, "phandle", PHANDLE_XICP)));
464 
465     _FDT((fdt_end_node(fdt)));
466 
467     /* vdevice */
468     _FDT((fdt_begin_node(fdt, "vdevice")));
469 
470     _FDT((fdt_property_string(fdt, "device_type", "vdevice")));
471     _FDT((fdt_property_string(fdt, "compatible", "IBM,vdevice")));
472     _FDT((fdt_property_cell(fdt, "#address-cells", 0x1)));
473     _FDT((fdt_property_cell(fdt, "#size-cells", 0x0)));
474     _FDT((fdt_property_cell(fdt, "#interrupt-cells", 0x2)));
475     _FDT((fdt_property(fdt, "interrupt-controller", NULL, 0)));
476 
477     _FDT((fdt_end_node(fdt)));
478 
479     /* event-sources */
480     spapr_events_fdt_skel(fdt, epow_irq);
481 
482     /* /hypervisor node */
483     if (kvm_enabled()) {
484         uint8_t hypercall[16];
485 
486         /* indicate KVM hypercall interface */
487         _FDT((fdt_begin_node(fdt, "hypervisor")));
488         _FDT((fdt_property_string(fdt, "compatible", "linux,kvm")));
489         if (kvmppc_has_cap_fixup_hcalls()) {
490             /*
491              * Older KVM versions with older guest kernels were broken with the
492              * magic page, don't allow the guest to map it.
493              */
494             kvmppc_get_hypercall(first_cpu->env_ptr, hypercall,
495                                  sizeof(hypercall));
496             _FDT((fdt_property(fdt, "hcall-instructions", hypercall,
497                               sizeof(hypercall))));
498         }
499         _FDT((fdt_end_node(fdt)));
500     }
501 
502     _FDT((fdt_end_node(fdt))); /* close root node */
503     _FDT((fdt_finish(fdt)));
504 
505     return fdt;
506 }
507 
508 static int spapr_populate_memory_node(void *fdt, int nodeid, hwaddr start,
509                                        hwaddr size)
510 {
511     uint32_t associativity[] = {
512         cpu_to_be32(0x4), /* length */
513         cpu_to_be32(0x0), cpu_to_be32(0x0),
514         cpu_to_be32(0x0), cpu_to_be32(nodeid)
515     };
516     char mem_name[32];
517     uint64_t mem_reg_property[2];
518     int off;
519 
520     mem_reg_property[0] = cpu_to_be64(start);
521     mem_reg_property[1] = cpu_to_be64(size);
522 
523     sprintf(mem_name, "memory@" TARGET_FMT_lx, start);
524     off = fdt_add_subnode(fdt, 0, mem_name);
525     _FDT(off);
526     _FDT((fdt_setprop_string(fdt, off, "device_type", "memory")));
527     _FDT((fdt_setprop(fdt, off, "reg", mem_reg_property,
528                       sizeof(mem_reg_property))));
529     _FDT((fdt_setprop(fdt, off, "ibm,associativity", associativity,
530                       sizeof(associativity))));
531     return off;
532 }
533 
534 static int spapr_populate_memory(sPAPRMachineState *spapr, void *fdt)
535 {
536     MachineState *machine = MACHINE(spapr);
537     hwaddr mem_start, node_size;
538     int i, nb_nodes = nb_numa_nodes;
539     NodeInfo *nodes = numa_info;
540     NodeInfo ramnode;
541 
542     /* No NUMA nodes, assume there is just one node with whole RAM */
543     if (!nb_numa_nodes) {
544         nb_nodes = 1;
545         ramnode.node_mem = machine->ram_size;
546         nodes = &ramnode;
547     }
548 
549     for (i = 0, mem_start = 0; i < nb_nodes; ++i) {
550         if (!nodes[i].node_mem) {
551             continue;
552         }
553         if (mem_start >= machine->ram_size) {
554             node_size = 0;
555         } else {
556             node_size = nodes[i].node_mem;
557             if (node_size > machine->ram_size - mem_start) {
558                 node_size = machine->ram_size - mem_start;
559             }
560         }
561         if (!mem_start) {
562             /* ppc_spapr_init() checks for rma_size <= node0_size already */
563             spapr_populate_memory_node(fdt, i, 0, spapr->rma_size);
564             mem_start += spapr->rma_size;
565             node_size -= spapr->rma_size;
566         }
567         for ( ; node_size; ) {
568             hwaddr sizetmp = pow2floor(node_size);
569 
570             /* mem_start != 0 here */
571             if (ctzl(mem_start) < ctzl(sizetmp)) {
572                 sizetmp = 1ULL << ctzl(mem_start);
573             }
574 
575             spapr_populate_memory_node(fdt, i, mem_start, sizetmp);
576             node_size -= sizetmp;
577             mem_start += sizetmp;
578         }
579     }
580 
581     return 0;
582 }
583 
584 static void spapr_populate_cpu_dt(CPUState *cs, void *fdt, int offset,
585                                   sPAPRMachineState *spapr)
586 {
587     PowerPCCPU *cpu = POWERPC_CPU(cs);
588     CPUPPCState *env = &cpu->env;
589     PowerPCCPUClass *pcc = POWERPC_CPU_GET_CLASS(cs);
590     int index = ppc_get_vcpu_dt_id(cpu);
591     uint32_t segs[] = {cpu_to_be32(28), cpu_to_be32(40),
592                        0xffffffff, 0xffffffff};
593     uint32_t tbfreq = kvm_enabled() ? kvmppc_get_tbfreq() : TIMEBASE_FREQ;
594     uint32_t cpufreq = kvm_enabled() ? kvmppc_get_clockfreq() : 1000000000;
595     uint32_t page_sizes_prop[64];
596     size_t page_sizes_prop_size;
597     uint32_t vcpus_per_socket = smp_threads * smp_cores;
598     uint32_t pft_size_prop[] = {0, cpu_to_be32(spapr->htab_shift)};
599 
600     /* Note: we keep CI large pages off for now because a 64K capable guest
601      * provisioned with large pages might otherwise try to map a qemu
602      * framebuffer (or other kind of memory mapped PCI BAR) using 64K pages
603      * even if that qemu runs on a 4k host.
604      *
605      * We can later add this bit back when we are confident this is not
606      * an issue (!HV KVM or 64K host)
607      */
608     uint8_t pa_features_206[] = { 6, 0,
609         0xf6, 0x1f, 0xc7, 0x00, 0x80, 0xc0 };
610     uint8_t pa_features_207[] = { 24, 0,
611         0xf6, 0x1f, 0xc7, 0xc0, 0x80, 0xf0,
612         0x80, 0x00, 0x00, 0x00, 0x00, 0x00,
613         0x00, 0x00, 0x00, 0x00, 0x80, 0x00,
614         0x80, 0x00, 0x80, 0x00, 0x80, 0x00 };
615     uint8_t *pa_features;
616     size_t pa_size;
617 
618     _FDT((fdt_setprop_cell(fdt, offset, "reg", index)));
619     _FDT((fdt_setprop_string(fdt, offset, "device_type", "cpu")));
620 
621     _FDT((fdt_setprop_cell(fdt, offset, "cpu-version", env->spr[SPR_PVR])));
622     _FDT((fdt_setprop_cell(fdt, offset, "d-cache-block-size",
623                            env->dcache_line_size)));
624     _FDT((fdt_setprop_cell(fdt, offset, "d-cache-line-size",
625                            env->dcache_line_size)));
626     _FDT((fdt_setprop_cell(fdt, offset, "i-cache-block-size",
627                            env->icache_line_size)));
628     _FDT((fdt_setprop_cell(fdt, offset, "i-cache-line-size",
629                            env->icache_line_size)));
630 
631     if (pcc->l1_dcache_size) {
632         _FDT((fdt_setprop_cell(fdt, offset, "d-cache-size",
633                                pcc->l1_dcache_size)));
634     } else {
635         fprintf(stderr, "Warning: Unknown L1 dcache size for cpu\n");
636     }
637     if (pcc->l1_icache_size) {
638         _FDT((fdt_setprop_cell(fdt, offset, "i-cache-size",
639                                pcc->l1_icache_size)));
640     } else {
641         fprintf(stderr, "Warning: Unknown L1 icache size for cpu\n");
642     }
643 
644     _FDT((fdt_setprop_cell(fdt, offset, "timebase-frequency", tbfreq)));
645     _FDT((fdt_setprop_cell(fdt, offset, "clock-frequency", cpufreq)));
646     _FDT((fdt_setprop_cell(fdt, offset, "slb-size", env->slb_nr)));
647     _FDT((fdt_setprop_cell(fdt, offset, "ibm,slb-size", env->slb_nr)));
648     _FDT((fdt_setprop_string(fdt, offset, "status", "okay")));
649     _FDT((fdt_setprop(fdt, offset, "64-bit", NULL, 0)));
650 
651     if (env->spr_cb[SPR_PURR].oea_read) {
652         _FDT((fdt_setprop(fdt, offset, "ibm,purr", NULL, 0)));
653     }
654 
655     if (env->mmu_model & POWERPC_MMU_1TSEG) {
656         _FDT((fdt_setprop(fdt, offset, "ibm,processor-segment-sizes",
657                           segs, sizeof(segs))));
658     }
659 
660     /* Advertise VMX/VSX (vector extensions) if available
661      *   0 / no property == no vector extensions
662      *   1               == VMX / Altivec available
663      *   2               == VSX available */
664     if (env->insns_flags & PPC_ALTIVEC) {
665         uint32_t vmx = (env->insns_flags2 & PPC2_VSX) ? 2 : 1;
666 
667         _FDT((fdt_setprop_cell(fdt, offset, "ibm,vmx", vmx)));
668     }
669 
670     /* Advertise DFP (Decimal Floating Point) if available
671      *   0 / no property == no DFP
672      *   1               == DFP available */
673     if (env->insns_flags2 & PPC2_DFP) {
674         _FDT((fdt_setprop_cell(fdt, offset, "ibm,dfp", 1)));
675     }
676 
677     page_sizes_prop_size = create_page_sizes_prop(env, page_sizes_prop,
678                                                   sizeof(page_sizes_prop));
679     if (page_sizes_prop_size) {
680         _FDT((fdt_setprop(fdt, offset, "ibm,segment-page-sizes",
681                           page_sizes_prop, page_sizes_prop_size)));
682     }
683 
684     /* Do the ibm,pa-features property, adjust it for ci-large-pages */
685     if (env->mmu_model == POWERPC_MMU_2_06) {
686         pa_features = pa_features_206;
687         pa_size = sizeof(pa_features_206);
688     } else /* env->mmu_model == POWERPC_MMU_2_07 */ {
689         pa_features = pa_features_207;
690         pa_size = sizeof(pa_features_207);
691     }
692     if (env->ci_large_pages) {
693         pa_features[3] |= 0x20;
694     }
695     _FDT((fdt_setprop(fdt, offset, "ibm,pa-features", pa_features, pa_size)));
696 
697     _FDT((fdt_setprop_cell(fdt, offset, "ibm,chip-id",
698                            cs->cpu_index / vcpus_per_socket)));
699 
700     _FDT((fdt_setprop(fdt, offset, "ibm,pft-size",
701                       pft_size_prop, sizeof(pft_size_prop))));
702 
703     _FDT(spapr_fixup_cpu_numa_dt(fdt, offset, cs));
704 
705     _FDT(spapr_fixup_cpu_smt_dt(fdt, offset, cpu,
706                                 ppc_get_compat_smt_threads(cpu)));
707 }
708 
709 static void spapr_populate_cpus_dt_node(void *fdt, sPAPRMachineState *spapr)
710 {
711     CPUState *cs;
712     int cpus_offset;
713     char *nodename;
714     int smt = kvmppc_smt_threads();
715 
716     cpus_offset = fdt_add_subnode(fdt, 0, "cpus");
717     _FDT(cpus_offset);
718     _FDT((fdt_setprop_cell(fdt, cpus_offset, "#address-cells", 0x1)));
719     _FDT((fdt_setprop_cell(fdt, cpus_offset, "#size-cells", 0x0)));
720 
721     /*
722      * We walk the CPUs in reverse order to ensure that CPU DT nodes
723      * created by fdt_add_subnode() end up in the right order in FDT
724      * for the guest kernel the enumerate the CPUs correctly.
725      */
726     CPU_FOREACH_REVERSE(cs) {
727         PowerPCCPU *cpu = POWERPC_CPU(cs);
728         int index = ppc_get_vcpu_dt_id(cpu);
729         DeviceClass *dc = DEVICE_GET_CLASS(cs);
730         int offset;
731 
732         if ((index % smt) != 0) {
733             continue;
734         }
735 
736         nodename = g_strdup_printf("%s@%x", dc->fw_name, index);
737         offset = fdt_add_subnode(fdt, cpus_offset, nodename);
738         g_free(nodename);
739         _FDT(offset);
740         spapr_populate_cpu_dt(cs, fdt, offset, spapr);
741     }
742 
743 }
744 
745 /*
746  * Adds ibm,dynamic-reconfiguration-memory node.
747  * Refer to docs/specs/ppc-spapr-hotplug.txt for the documentation
748  * of this device tree node.
749  */
750 static int spapr_populate_drconf_memory(sPAPRMachineState *spapr, void *fdt)
751 {
752     MachineState *machine = MACHINE(spapr);
753     int ret, i, offset;
754     uint64_t lmb_size = SPAPR_MEMORY_BLOCK_SIZE;
755     uint32_t prop_lmb_size[] = {0, cpu_to_be32(lmb_size)};
756     uint32_t nr_lmbs = (machine->maxram_size - machine->ram_size)/lmb_size;
757     uint32_t *int_buf, *cur_index, buf_len;
758     int nr_nodes = nb_numa_nodes ? nb_numa_nodes : 1;
759 
760     /*
761      * Allocate enough buffer size to fit in ibm,dynamic-memory
762      * or ibm,associativity-lookup-arrays
763      */
764     buf_len = MAX(nr_lmbs * SPAPR_DR_LMB_LIST_ENTRY_SIZE + 1, nr_nodes * 4 + 2)
765               * sizeof(uint32_t);
766     cur_index = int_buf = g_malloc0(buf_len);
767 
768     offset = fdt_add_subnode(fdt, 0, "ibm,dynamic-reconfiguration-memory");
769 
770     ret = fdt_setprop(fdt, offset, "ibm,lmb-size", prop_lmb_size,
771                     sizeof(prop_lmb_size));
772     if (ret < 0) {
773         goto out;
774     }
775 
776     ret = fdt_setprop_cell(fdt, offset, "ibm,memory-flags-mask", 0xff);
777     if (ret < 0) {
778         goto out;
779     }
780 
781     ret = fdt_setprop_cell(fdt, offset, "ibm,memory-preservation-time", 0x0);
782     if (ret < 0) {
783         goto out;
784     }
785 
786     /* ibm,dynamic-memory */
787     int_buf[0] = cpu_to_be32(nr_lmbs);
788     cur_index++;
789     for (i = 0; i < nr_lmbs; i++) {
790         sPAPRDRConnector *drc;
791         sPAPRDRConnectorClass *drck;
792         uint64_t addr = i * lmb_size + spapr->hotplug_memory.base;;
793         uint32_t *dynamic_memory = cur_index;
794 
795         drc = spapr_dr_connector_by_id(SPAPR_DR_CONNECTOR_TYPE_LMB,
796                                        addr/lmb_size);
797         g_assert(drc);
798         drck = SPAPR_DR_CONNECTOR_GET_CLASS(drc);
799 
800         dynamic_memory[0] = cpu_to_be32(addr >> 32);
801         dynamic_memory[1] = cpu_to_be32(addr & 0xffffffff);
802         dynamic_memory[2] = cpu_to_be32(drck->get_index(drc));
803         dynamic_memory[3] = cpu_to_be32(0); /* reserved */
804         dynamic_memory[4] = cpu_to_be32(numa_get_node(addr, NULL));
805         if (addr < machine->ram_size ||
806                     memory_region_present(get_system_memory(), addr)) {
807             dynamic_memory[5] = cpu_to_be32(SPAPR_LMB_FLAGS_ASSIGNED);
808         } else {
809             dynamic_memory[5] = cpu_to_be32(0);
810         }
811 
812         cur_index += SPAPR_DR_LMB_LIST_ENTRY_SIZE;
813     }
814     ret = fdt_setprop(fdt, offset, "ibm,dynamic-memory", int_buf, buf_len);
815     if (ret < 0) {
816         goto out;
817     }
818 
819     /* ibm,associativity-lookup-arrays */
820     cur_index = int_buf;
821     int_buf[0] = cpu_to_be32(nr_nodes);
822     int_buf[1] = cpu_to_be32(4); /* Number of entries per associativity list */
823     cur_index += 2;
824     for (i = 0; i < nr_nodes; i++) {
825         uint32_t associativity[] = {
826             cpu_to_be32(0x0),
827             cpu_to_be32(0x0),
828             cpu_to_be32(0x0),
829             cpu_to_be32(i)
830         };
831         memcpy(cur_index, associativity, sizeof(associativity));
832         cur_index += 4;
833     }
834     ret = fdt_setprop(fdt, offset, "ibm,associativity-lookup-arrays", int_buf,
835             (cur_index - int_buf) * sizeof(uint32_t));
836 out:
837     g_free(int_buf);
838     return ret;
839 }
840 
841 int spapr_h_cas_compose_response(sPAPRMachineState *spapr,
842                                  target_ulong addr, target_ulong size,
843                                  bool cpu_update, bool memory_update)
844 {
845     void *fdt, *fdt_skel;
846     sPAPRDeviceTreeUpdateHeader hdr = { .version_id = 1 };
847     sPAPRMachineClass *smc = SPAPR_MACHINE_GET_CLASS(qdev_get_machine());
848 
849     size -= sizeof(hdr);
850 
851     /* Create sceleton */
852     fdt_skel = g_malloc0(size);
853     _FDT((fdt_create(fdt_skel, size)));
854     _FDT((fdt_begin_node(fdt_skel, "")));
855     _FDT((fdt_end_node(fdt_skel)));
856     _FDT((fdt_finish(fdt_skel)));
857     fdt = g_malloc0(size);
858     _FDT((fdt_open_into(fdt_skel, fdt, size)));
859     g_free(fdt_skel);
860 
861     /* Fixup cpu nodes */
862     if (cpu_update) {
863         _FDT((spapr_fixup_cpu_dt(fdt, spapr)));
864     }
865 
866     /* Generate memory nodes or ibm,dynamic-reconfiguration-memory node */
867     if (memory_update && smc->dr_lmb_enabled) {
868         _FDT((spapr_populate_drconf_memory(spapr, fdt)));
869     }
870 
871     /* Pack resulting tree */
872     _FDT((fdt_pack(fdt)));
873 
874     if (fdt_totalsize(fdt) + sizeof(hdr) > size) {
875         trace_spapr_cas_failed(size);
876         return -1;
877     }
878 
879     cpu_physical_memory_write(addr, &hdr, sizeof(hdr));
880     cpu_physical_memory_write(addr + sizeof(hdr), fdt, fdt_totalsize(fdt));
881     trace_spapr_cas_continue(fdt_totalsize(fdt) + sizeof(hdr));
882     g_free(fdt);
883 
884     return 0;
885 }
886 
887 static void spapr_finalize_fdt(sPAPRMachineState *spapr,
888                                hwaddr fdt_addr,
889                                hwaddr rtas_addr,
890                                hwaddr rtas_size)
891 {
892     MachineState *machine = MACHINE(qdev_get_machine());
893     sPAPRMachineClass *smc = SPAPR_MACHINE_GET_CLASS(machine);
894     const char *boot_device = machine->boot_order;
895     int ret, i;
896     size_t cb = 0;
897     char *bootlist;
898     void *fdt;
899     sPAPRPHBState *phb;
900 
901     fdt = g_malloc(FDT_MAX_SIZE);
902 
903     /* open out the base tree into a temp buffer for the final tweaks */
904     _FDT((fdt_open_into(spapr->fdt_skel, fdt, FDT_MAX_SIZE)));
905 
906     ret = spapr_populate_memory(spapr, fdt);
907     if (ret < 0) {
908         fprintf(stderr, "couldn't setup memory nodes in fdt\n");
909         exit(1);
910     }
911 
912     ret = spapr_populate_vdevice(spapr->vio_bus, fdt);
913     if (ret < 0) {
914         fprintf(stderr, "couldn't setup vio devices in fdt\n");
915         exit(1);
916     }
917 
918     if (object_resolve_path_type("", TYPE_SPAPR_RNG, NULL)) {
919         ret = spapr_rng_populate_dt(fdt);
920         if (ret < 0) {
921             fprintf(stderr, "could not set up rng device in the fdt\n");
922             exit(1);
923         }
924     }
925 
926     QLIST_FOREACH(phb, &spapr->phbs, list) {
927         ret = spapr_populate_pci_dt(phb, PHANDLE_XICP, fdt);
928     }
929 
930     if (ret < 0) {
931         fprintf(stderr, "couldn't setup PCI devices in fdt\n");
932         exit(1);
933     }
934 
935     /* RTAS */
936     ret = spapr_rtas_device_tree_setup(fdt, rtas_addr, rtas_size);
937     if (ret < 0) {
938         fprintf(stderr, "Couldn't set up RTAS device tree properties\n");
939     }
940 
941     /* cpus */
942     spapr_populate_cpus_dt_node(fdt, spapr);
943 
944     bootlist = get_boot_devices_list(&cb, true);
945     if (cb && bootlist) {
946         int offset = fdt_path_offset(fdt, "/chosen");
947         if (offset < 0) {
948             exit(1);
949         }
950         for (i = 0; i < cb; i++) {
951             if (bootlist[i] == '\n') {
952                 bootlist[i] = ' ';
953             }
954 
955         }
956         ret = fdt_setprop_string(fdt, offset, "qemu,boot-list", bootlist);
957     }
958 
959     if (boot_device && strlen(boot_device)) {
960         int offset = fdt_path_offset(fdt, "/chosen");
961 
962         if (offset < 0) {
963             exit(1);
964         }
965         fdt_setprop_string(fdt, offset, "qemu,boot-device", boot_device);
966     }
967 
968     if (!spapr->has_graphics) {
969         spapr_populate_chosen_stdout(fdt, spapr->vio_bus);
970     }
971 
972     if (smc->dr_lmb_enabled) {
973         _FDT(spapr_drc_populate_dt(fdt, 0, NULL, SPAPR_DR_CONNECTOR_TYPE_LMB));
974     }
975 
976     _FDT((fdt_pack(fdt)));
977 
978     if (fdt_totalsize(fdt) > FDT_MAX_SIZE) {
979         error_report("FDT too big ! 0x%x bytes (max is 0x%x)",
980                      fdt_totalsize(fdt), FDT_MAX_SIZE);
981         exit(1);
982     }
983 
984     qemu_fdt_dumpdtb(fdt, fdt_totalsize(fdt));
985     cpu_physical_memory_write(fdt_addr, fdt, fdt_totalsize(fdt));
986 
987     g_free(bootlist);
988     g_free(fdt);
989 }
990 
991 static uint64_t translate_kernel_address(void *opaque, uint64_t addr)
992 {
993     return (addr & 0x0fffffff) + KERNEL_LOAD_ADDR;
994 }
995 
996 static void emulate_spapr_hypercall(PowerPCCPU *cpu)
997 {
998     CPUPPCState *env = &cpu->env;
999 
1000     if (msr_pr) {
1001         hcall_dprintf("Hypercall made with MSR[PR]=1\n");
1002         env->gpr[3] = H_PRIVILEGE;
1003     } else {
1004         env->gpr[3] = spapr_hypercall(cpu, env->gpr[3], &env->gpr[4]);
1005     }
1006 }
1007 
1008 #define HPTE(_table, _i)   (void *)(((uint64_t *)(_table)) + ((_i) * 2))
1009 #define HPTE_VALID(_hpte)  (tswap64(*((uint64_t *)(_hpte))) & HPTE64_V_VALID)
1010 #define HPTE_DIRTY(_hpte)  (tswap64(*((uint64_t *)(_hpte))) & HPTE64_V_HPTE_DIRTY)
1011 #define CLEAN_HPTE(_hpte)  ((*(uint64_t *)(_hpte)) &= tswap64(~HPTE64_V_HPTE_DIRTY))
1012 #define DIRTY_HPTE(_hpte)  ((*(uint64_t *)(_hpte)) |= tswap64(HPTE64_V_HPTE_DIRTY))
1013 
1014 static void spapr_alloc_htab(sPAPRMachineState *spapr)
1015 {
1016     long shift;
1017     int index;
1018 
1019     /* allocate hash page table.  For now we always make this 16mb,
1020      * later we should probably make it scale to the size of guest
1021      * RAM */
1022 
1023     shift = kvmppc_reset_htab(spapr->htab_shift);
1024     if (shift < 0) {
1025         /*
1026          * For HV KVM, host kernel will return -ENOMEM when requested
1027          * HTAB size can't be allocated.
1028          */
1029         error_setg(&error_abort, "Failed to allocate HTAB of requested size, try with smaller maxmem");
1030     } else if (shift > 0) {
1031         /*
1032          * Kernel handles htab, we don't need to allocate one
1033          *
1034          * Older kernels can fall back to lower HTAB shift values,
1035          * but we don't allow booting of such guests.
1036          */
1037         if (shift != spapr->htab_shift) {
1038             error_setg(&error_abort, "Failed to allocate HTAB of requested size, try with smaller maxmem");
1039         }
1040 
1041         spapr->htab_shift = shift;
1042         kvmppc_kern_htab = true;
1043     } else {
1044         /* Allocate htab */
1045         spapr->htab = qemu_memalign(HTAB_SIZE(spapr), HTAB_SIZE(spapr));
1046 
1047         /* And clear it */
1048         memset(spapr->htab, 0, HTAB_SIZE(spapr));
1049 
1050         for (index = 0; index < HTAB_SIZE(spapr) / HASH_PTE_SIZE_64; index++) {
1051             DIRTY_HPTE(HPTE(spapr->htab, index));
1052         }
1053     }
1054 }
1055 
1056 /*
1057  * Clear HTAB entries during reset.
1058  *
1059  * If host kernel has allocated HTAB, KVM_PPC_ALLOCATE_HTAB ioctl is
1060  * used to clear HTAB. Otherwise QEMU-allocated HTAB is cleared manually.
1061  */
1062 static void spapr_reset_htab(sPAPRMachineState *spapr)
1063 {
1064     long shift;
1065     int index;
1066 
1067     shift = kvmppc_reset_htab(spapr->htab_shift);
1068     if (shift < 0) {
1069         error_setg(&error_abort, "Failed to reset HTAB");
1070     } else if (shift > 0) {
1071         if (shift != spapr->htab_shift) {
1072             error_setg(&error_abort, "Requested HTAB allocation failed during reset");
1073         }
1074 
1075         /* Tell readers to update their file descriptor */
1076         if (spapr->htab_fd >= 0) {
1077             spapr->htab_fd_stale = true;
1078         }
1079     } else {
1080         memset(spapr->htab, 0, HTAB_SIZE(spapr));
1081 
1082         for (index = 0; index < HTAB_SIZE(spapr) / HASH_PTE_SIZE_64; index++) {
1083             DIRTY_HPTE(HPTE(spapr->htab, index));
1084         }
1085     }
1086 
1087     /* Update the RMA size if necessary */
1088     if (spapr->vrma_adjust) {
1089         spapr->rma_size = kvmppc_rma_size(spapr_node0_size(),
1090                                           spapr->htab_shift);
1091     }
1092 }
1093 
1094 static int find_unknown_sysbus_device(SysBusDevice *sbdev, void *opaque)
1095 {
1096     bool matched = false;
1097 
1098     if (object_dynamic_cast(OBJECT(sbdev), TYPE_SPAPR_PCI_HOST_BRIDGE)) {
1099         matched = true;
1100     }
1101 
1102     if (!matched) {
1103         error_report("Device %s is not supported by this machine yet.",
1104                      qdev_fw_name(DEVICE(sbdev)));
1105         exit(1);
1106     }
1107 
1108     return 0;
1109 }
1110 
1111 /*
1112  * A guest reset will cause spapr->htab_fd to become stale if being used.
1113  * Reopen the file descriptor to make sure the whole HTAB is properly read.
1114  */
1115 static int spapr_check_htab_fd(sPAPRMachineState *spapr)
1116 {
1117     int rc = 0;
1118 
1119     if (spapr->htab_fd_stale) {
1120         close(spapr->htab_fd);
1121         spapr->htab_fd = kvmppc_get_htab_fd(false);
1122         if (spapr->htab_fd < 0) {
1123             error_report("Unable to open fd for reading hash table from KVM: "
1124                          "%s", strerror(errno));
1125             rc = -1;
1126         }
1127         spapr->htab_fd_stale = false;
1128     }
1129 
1130     return rc;
1131 }
1132 
1133 static void ppc_spapr_reset(void)
1134 {
1135     sPAPRMachineState *spapr = SPAPR_MACHINE(qdev_get_machine());
1136     PowerPCCPU *first_ppc_cpu;
1137     uint32_t rtas_limit;
1138 
1139     /* Check for unknown sysbus devices */
1140     foreach_dynamic_sysbus_device(find_unknown_sysbus_device, NULL);
1141 
1142     /* Reset the hash table & recalc the RMA */
1143     spapr_reset_htab(spapr);
1144 
1145     qemu_devices_reset();
1146 
1147     /*
1148      * We place the device tree and RTAS just below either the top of the RMA,
1149      * or just below 2GB, whichever is lowere, so that it can be
1150      * processed with 32-bit real mode code if necessary
1151      */
1152     rtas_limit = MIN(spapr->rma_size, RTAS_MAX_ADDR);
1153     spapr->rtas_addr = rtas_limit - RTAS_MAX_SIZE;
1154     spapr->fdt_addr = spapr->rtas_addr - FDT_MAX_SIZE;
1155 
1156     /* Load the fdt */
1157     spapr_finalize_fdt(spapr, spapr->fdt_addr, spapr->rtas_addr,
1158                        spapr->rtas_size);
1159 
1160     /* Copy RTAS over */
1161     cpu_physical_memory_write(spapr->rtas_addr, spapr->rtas_blob,
1162                               spapr->rtas_size);
1163 
1164     /* Set up the entry state */
1165     first_ppc_cpu = POWERPC_CPU(first_cpu);
1166     first_ppc_cpu->env.gpr[3] = spapr->fdt_addr;
1167     first_ppc_cpu->env.gpr[5] = 0;
1168     first_cpu->halted = 0;
1169     first_ppc_cpu->env.nip = SPAPR_ENTRY_POINT;
1170 
1171 }
1172 
1173 static void spapr_cpu_reset(void *opaque)
1174 {
1175     sPAPRMachineState *spapr = SPAPR_MACHINE(qdev_get_machine());
1176     PowerPCCPU *cpu = opaque;
1177     CPUState *cs = CPU(cpu);
1178     CPUPPCState *env = &cpu->env;
1179 
1180     cpu_reset(cs);
1181 
1182     /* All CPUs start halted.  CPU0 is unhalted from the machine level
1183      * reset code and the rest are explicitly started up by the guest
1184      * using an RTAS call */
1185     cs->halted = 1;
1186 
1187     env->spr[SPR_HIOR] = 0;
1188 
1189     env->external_htab = (uint8_t *)spapr->htab;
1190     if (kvm_enabled() && !env->external_htab) {
1191         /*
1192          * HV KVM, set external_htab to 1 so our ppc_hash64_load_hpte*
1193          * functions do the right thing.
1194          */
1195         env->external_htab = (void *)1;
1196     }
1197     env->htab_base = -1;
1198     /*
1199      * htab_mask is the mask used to normalize hash value to PTEG index.
1200      * htab_shift is log2 of hash table size.
1201      * We have 8 hpte per group, and each hpte is 16 bytes.
1202      * ie have 128 bytes per hpte entry.
1203      */
1204     env->htab_mask = (1ULL << (spapr->htab_shift - 7)) - 1;
1205     env->spr[SPR_SDR1] = (target_ulong)(uintptr_t)spapr->htab |
1206         (spapr->htab_shift - 18);
1207 }
1208 
1209 static void spapr_create_nvram(sPAPRMachineState *spapr)
1210 {
1211     DeviceState *dev = qdev_create(&spapr->vio_bus->bus, "spapr-nvram");
1212     DriveInfo *dinfo = drive_get(IF_PFLASH, 0, 0);
1213 
1214     if (dinfo) {
1215         qdev_prop_set_drive_nofail(dev, "drive", blk_by_legacy_dinfo(dinfo));
1216     }
1217 
1218     qdev_init_nofail(dev);
1219 
1220     spapr->nvram = (struct sPAPRNVRAM *)dev;
1221 }
1222 
1223 static void spapr_rtc_create(sPAPRMachineState *spapr)
1224 {
1225     DeviceState *dev = qdev_create(NULL, TYPE_SPAPR_RTC);
1226 
1227     qdev_init_nofail(dev);
1228     spapr->rtc = dev;
1229 
1230     object_property_add_alias(qdev_get_machine(), "rtc-time",
1231                               OBJECT(spapr->rtc), "date", NULL);
1232 }
1233 
1234 /* Returns whether we want to use VGA or not */
1235 static int spapr_vga_init(PCIBus *pci_bus)
1236 {
1237     switch (vga_interface_type) {
1238     case VGA_NONE:
1239         return false;
1240     case VGA_DEVICE:
1241         return true;
1242     case VGA_STD:
1243     case VGA_VIRTIO:
1244         return pci_vga_init(pci_bus) != NULL;
1245     default:
1246         fprintf(stderr, "This vga model is not supported,"
1247                 "currently it only supports -vga std\n");
1248         exit(0);
1249     }
1250 }
1251 
1252 static int spapr_post_load(void *opaque, int version_id)
1253 {
1254     sPAPRMachineState *spapr = (sPAPRMachineState *)opaque;
1255     int err = 0;
1256 
1257     /* In earlier versions, there was no separate qdev for the PAPR
1258      * RTC, so the RTC offset was stored directly in sPAPREnvironment.
1259      * So when migrating from those versions, poke the incoming offset
1260      * value into the RTC device */
1261     if (version_id < 3) {
1262         err = spapr_rtc_import_offset(spapr->rtc, spapr->rtc_offset);
1263     }
1264 
1265     return err;
1266 }
1267 
1268 static bool version_before_3(void *opaque, int version_id)
1269 {
1270     return version_id < 3;
1271 }
1272 
1273 static const VMStateDescription vmstate_spapr = {
1274     .name = "spapr",
1275     .version_id = 3,
1276     .minimum_version_id = 1,
1277     .post_load = spapr_post_load,
1278     .fields = (VMStateField[]) {
1279         /* used to be @next_irq */
1280         VMSTATE_UNUSED_BUFFER(version_before_3, 0, 4),
1281 
1282         /* RTC offset */
1283         VMSTATE_UINT64_TEST(rtc_offset, sPAPRMachineState, version_before_3),
1284 
1285         VMSTATE_PPC_TIMEBASE_V(tb, sPAPRMachineState, 2),
1286         VMSTATE_END_OF_LIST()
1287     },
1288 };
1289 
1290 static int htab_save_setup(QEMUFile *f, void *opaque)
1291 {
1292     sPAPRMachineState *spapr = opaque;
1293 
1294     /* "Iteration" header */
1295     qemu_put_be32(f, spapr->htab_shift);
1296 
1297     if (spapr->htab) {
1298         spapr->htab_save_index = 0;
1299         spapr->htab_first_pass = true;
1300     } else {
1301         assert(kvm_enabled());
1302 
1303         spapr->htab_fd = kvmppc_get_htab_fd(false);
1304         spapr->htab_fd_stale = false;
1305         if (spapr->htab_fd < 0) {
1306             fprintf(stderr, "Unable to open fd for reading hash table from KVM: %s\n",
1307                     strerror(errno));
1308             return -1;
1309         }
1310     }
1311 
1312 
1313     return 0;
1314 }
1315 
1316 static void htab_save_first_pass(QEMUFile *f, sPAPRMachineState *spapr,
1317                                  int64_t max_ns)
1318 {
1319     int htabslots = HTAB_SIZE(spapr) / HASH_PTE_SIZE_64;
1320     int index = spapr->htab_save_index;
1321     int64_t starttime = qemu_clock_get_ns(QEMU_CLOCK_REALTIME);
1322 
1323     assert(spapr->htab_first_pass);
1324 
1325     do {
1326         int chunkstart;
1327 
1328         /* Consume invalid HPTEs */
1329         while ((index < htabslots)
1330                && !HPTE_VALID(HPTE(spapr->htab, index))) {
1331             index++;
1332             CLEAN_HPTE(HPTE(spapr->htab, index));
1333         }
1334 
1335         /* Consume valid HPTEs */
1336         chunkstart = index;
1337         while ((index < htabslots) && (index - chunkstart < USHRT_MAX)
1338                && HPTE_VALID(HPTE(spapr->htab, index))) {
1339             index++;
1340             CLEAN_HPTE(HPTE(spapr->htab, index));
1341         }
1342 
1343         if (index > chunkstart) {
1344             int n_valid = index - chunkstart;
1345 
1346             qemu_put_be32(f, chunkstart);
1347             qemu_put_be16(f, n_valid);
1348             qemu_put_be16(f, 0);
1349             qemu_put_buffer(f, HPTE(spapr->htab, chunkstart),
1350                             HASH_PTE_SIZE_64 * n_valid);
1351 
1352             if ((qemu_clock_get_ns(QEMU_CLOCK_REALTIME) - starttime) > max_ns) {
1353                 break;
1354             }
1355         }
1356     } while ((index < htabslots) && !qemu_file_rate_limit(f));
1357 
1358     if (index >= htabslots) {
1359         assert(index == htabslots);
1360         index = 0;
1361         spapr->htab_first_pass = false;
1362     }
1363     spapr->htab_save_index = index;
1364 }
1365 
1366 static int htab_save_later_pass(QEMUFile *f, sPAPRMachineState *spapr,
1367                                 int64_t max_ns)
1368 {
1369     bool final = max_ns < 0;
1370     int htabslots = HTAB_SIZE(spapr) / HASH_PTE_SIZE_64;
1371     int examined = 0, sent = 0;
1372     int index = spapr->htab_save_index;
1373     int64_t starttime = qemu_clock_get_ns(QEMU_CLOCK_REALTIME);
1374 
1375     assert(!spapr->htab_first_pass);
1376 
1377     do {
1378         int chunkstart, invalidstart;
1379 
1380         /* Consume non-dirty HPTEs */
1381         while ((index < htabslots)
1382                && !HPTE_DIRTY(HPTE(spapr->htab, index))) {
1383             index++;
1384             examined++;
1385         }
1386 
1387         chunkstart = index;
1388         /* Consume valid dirty HPTEs */
1389         while ((index < htabslots) && (index - chunkstart < USHRT_MAX)
1390                && HPTE_DIRTY(HPTE(spapr->htab, index))
1391                && HPTE_VALID(HPTE(spapr->htab, index))) {
1392             CLEAN_HPTE(HPTE(spapr->htab, index));
1393             index++;
1394             examined++;
1395         }
1396 
1397         invalidstart = index;
1398         /* Consume invalid dirty HPTEs */
1399         while ((index < htabslots) && (index - invalidstart < USHRT_MAX)
1400                && HPTE_DIRTY(HPTE(spapr->htab, index))
1401                && !HPTE_VALID(HPTE(spapr->htab, index))) {
1402             CLEAN_HPTE(HPTE(spapr->htab, index));
1403             index++;
1404             examined++;
1405         }
1406 
1407         if (index > chunkstart) {
1408             int n_valid = invalidstart - chunkstart;
1409             int n_invalid = index - invalidstart;
1410 
1411             qemu_put_be32(f, chunkstart);
1412             qemu_put_be16(f, n_valid);
1413             qemu_put_be16(f, n_invalid);
1414             qemu_put_buffer(f, HPTE(spapr->htab, chunkstart),
1415                             HASH_PTE_SIZE_64 * n_valid);
1416             sent += index - chunkstart;
1417 
1418             if (!final && (qemu_clock_get_ns(QEMU_CLOCK_REALTIME) - starttime) > max_ns) {
1419                 break;
1420             }
1421         }
1422 
1423         if (examined >= htabslots) {
1424             break;
1425         }
1426 
1427         if (index >= htabslots) {
1428             assert(index == htabslots);
1429             index = 0;
1430         }
1431     } while ((examined < htabslots) && (!qemu_file_rate_limit(f) || final));
1432 
1433     if (index >= htabslots) {
1434         assert(index == htabslots);
1435         index = 0;
1436     }
1437 
1438     spapr->htab_save_index = index;
1439 
1440     return (examined >= htabslots) && (sent == 0) ? 1 : 0;
1441 }
1442 
1443 #define MAX_ITERATION_NS    5000000 /* 5 ms */
1444 #define MAX_KVM_BUF_SIZE    2048
1445 
1446 static int htab_save_iterate(QEMUFile *f, void *opaque)
1447 {
1448     sPAPRMachineState *spapr = opaque;
1449     int rc = 0;
1450 
1451     /* Iteration header */
1452     qemu_put_be32(f, 0);
1453 
1454     if (!spapr->htab) {
1455         assert(kvm_enabled());
1456 
1457         rc = spapr_check_htab_fd(spapr);
1458         if (rc < 0) {
1459             return rc;
1460         }
1461 
1462         rc = kvmppc_save_htab(f, spapr->htab_fd,
1463                               MAX_KVM_BUF_SIZE, MAX_ITERATION_NS);
1464         if (rc < 0) {
1465             return rc;
1466         }
1467     } else  if (spapr->htab_first_pass) {
1468         htab_save_first_pass(f, spapr, MAX_ITERATION_NS);
1469     } else {
1470         rc = htab_save_later_pass(f, spapr, MAX_ITERATION_NS);
1471     }
1472 
1473     /* End marker */
1474     qemu_put_be32(f, 0);
1475     qemu_put_be16(f, 0);
1476     qemu_put_be16(f, 0);
1477 
1478     return rc;
1479 }
1480 
1481 static int htab_save_complete(QEMUFile *f, void *opaque)
1482 {
1483     sPAPRMachineState *spapr = opaque;
1484 
1485     /* Iteration header */
1486     qemu_put_be32(f, 0);
1487 
1488     if (!spapr->htab) {
1489         int rc;
1490 
1491         assert(kvm_enabled());
1492 
1493         rc = spapr_check_htab_fd(spapr);
1494         if (rc < 0) {
1495             return rc;
1496         }
1497 
1498         rc = kvmppc_save_htab(f, spapr->htab_fd, MAX_KVM_BUF_SIZE, -1);
1499         if (rc < 0) {
1500             return rc;
1501         }
1502         close(spapr->htab_fd);
1503         spapr->htab_fd = -1;
1504     } else {
1505         htab_save_later_pass(f, spapr, -1);
1506     }
1507 
1508     /* End marker */
1509     qemu_put_be32(f, 0);
1510     qemu_put_be16(f, 0);
1511     qemu_put_be16(f, 0);
1512 
1513     return 0;
1514 }
1515 
1516 static int htab_load(QEMUFile *f, void *opaque, int version_id)
1517 {
1518     sPAPRMachineState *spapr = opaque;
1519     uint32_t section_hdr;
1520     int fd = -1;
1521 
1522     if (version_id < 1 || version_id > 1) {
1523         fprintf(stderr, "htab_load() bad version\n");
1524         return -EINVAL;
1525     }
1526 
1527     section_hdr = qemu_get_be32(f);
1528 
1529     if (section_hdr) {
1530         /* First section, just the hash shift */
1531         if (spapr->htab_shift != section_hdr) {
1532             error_report("htab_shift mismatch: source %d target %d",
1533                          section_hdr, spapr->htab_shift);
1534             return -EINVAL;
1535         }
1536         return 0;
1537     }
1538 
1539     if (!spapr->htab) {
1540         assert(kvm_enabled());
1541 
1542         fd = kvmppc_get_htab_fd(true);
1543         if (fd < 0) {
1544             fprintf(stderr, "Unable to open fd to restore KVM hash table: %s\n",
1545                     strerror(errno));
1546         }
1547     }
1548 
1549     while (true) {
1550         uint32_t index;
1551         uint16_t n_valid, n_invalid;
1552 
1553         index = qemu_get_be32(f);
1554         n_valid = qemu_get_be16(f);
1555         n_invalid = qemu_get_be16(f);
1556 
1557         if ((index == 0) && (n_valid == 0) && (n_invalid == 0)) {
1558             /* End of Stream */
1559             break;
1560         }
1561 
1562         if ((index + n_valid + n_invalid) >
1563             (HTAB_SIZE(spapr) / HASH_PTE_SIZE_64)) {
1564             /* Bad index in stream */
1565             fprintf(stderr, "htab_load() bad index %d (%hd+%hd entries) "
1566                     "in htab stream (htab_shift=%d)\n", index, n_valid, n_invalid,
1567                     spapr->htab_shift);
1568             return -EINVAL;
1569         }
1570 
1571         if (spapr->htab) {
1572             if (n_valid) {
1573                 qemu_get_buffer(f, HPTE(spapr->htab, index),
1574                                 HASH_PTE_SIZE_64 * n_valid);
1575             }
1576             if (n_invalid) {
1577                 memset(HPTE(spapr->htab, index + n_valid), 0,
1578                        HASH_PTE_SIZE_64 * n_invalid);
1579             }
1580         } else {
1581             int rc;
1582 
1583             assert(fd >= 0);
1584 
1585             rc = kvmppc_load_htab_chunk(f, fd, index, n_valid, n_invalid);
1586             if (rc < 0) {
1587                 return rc;
1588             }
1589         }
1590     }
1591 
1592     if (!spapr->htab) {
1593         assert(fd >= 0);
1594         close(fd);
1595     }
1596 
1597     return 0;
1598 }
1599 
1600 static SaveVMHandlers savevm_htab_handlers = {
1601     .save_live_setup = htab_save_setup,
1602     .save_live_iterate = htab_save_iterate,
1603     .save_live_complete_precopy = htab_save_complete,
1604     .load_state = htab_load,
1605 };
1606 
1607 static void spapr_boot_set(void *opaque, const char *boot_device,
1608                            Error **errp)
1609 {
1610     MachineState *machine = MACHINE(qdev_get_machine());
1611     machine->boot_order = g_strdup(boot_device);
1612 }
1613 
1614 static void spapr_cpu_init(sPAPRMachineState *spapr, PowerPCCPU *cpu)
1615 {
1616     CPUPPCState *env = &cpu->env;
1617 
1618     /* Set time-base frequency to 512 MHz */
1619     cpu_ppc_tb_init(env, TIMEBASE_FREQ);
1620 
1621     /* PAPR always has exception vectors in RAM not ROM. To ensure this,
1622      * MSR[IP] should never be set.
1623      */
1624     env->msr_mask &= ~(1 << 6);
1625 
1626     /* Tell KVM that we're in PAPR mode */
1627     if (kvm_enabled()) {
1628         kvmppc_set_papr(cpu);
1629     }
1630 
1631     if (cpu->max_compat) {
1632         if (ppc_set_compat(cpu, cpu->max_compat) < 0) {
1633             exit(1);
1634         }
1635     }
1636 
1637     xics_cpu_setup(spapr->icp, cpu);
1638 
1639     qemu_register_reset(spapr_cpu_reset, cpu);
1640 }
1641 
1642 /*
1643  * Reset routine for LMB DR devices.
1644  *
1645  * Unlike PCI DR devices, LMB DR devices explicitly register this reset
1646  * routine. Reset for PCI DR devices will be handled by PHB reset routine
1647  * when it walks all its children devices. LMB devices reset occurs
1648  * as part of spapr_ppc_reset().
1649  */
1650 static void spapr_drc_reset(void *opaque)
1651 {
1652     sPAPRDRConnector *drc = opaque;
1653     DeviceState *d = DEVICE(drc);
1654 
1655     if (d) {
1656         device_reset(d);
1657     }
1658 }
1659 
1660 static void spapr_create_lmb_dr_connectors(sPAPRMachineState *spapr)
1661 {
1662     MachineState *machine = MACHINE(spapr);
1663     uint64_t lmb_size = SPAPR_MEMORY_BLOCK_SIZE;
1664     uint32_t nr_lmbs = (machine->maxram_size - machine->ram_size)/lmb_size;
1665     int i;
1666 
1667     for (i = 0; i < nr_lmbs; i++) {
1668         sPAPRDRConnector *drc;
1669         uint64_t addr;
1670 
1671         addr = i * lmb_size + spapr->hotplug_memory.base;
1672         drc = spapr_dr_connector_new(OBJECT(spapr), SPAPR_DR_CONNECTOR_TYPE_LMB,
1673                                      addr/lmb_size);
1674         qemu_register_reset(spapr_drc_reset, drc);
1675     }
1676 }
1677 
1678 /*
1679  * If RAM size, maxmem size and individual node mem sizes aren't aligned
1680  * to SPAPR_MEMORY_BLOCK_SIZE(256MB), then refuse to start the guest
1681  * since we can't support such unaligned sizes with DRCONF_MEMORY.
1682  */
1683 static void spapr_validate_node_memory(MachineState *machine)
1684 {
1685     int i;
1686 
1687     if (machine->maxram_size % SPAPR_MEMORY_BLOCK_SIZE ||
1688         machine->ram_size % SPAPR_MEMORY_BLOCK_SIZE) {
1689         error_report("Can't support memory configuration where RAM size "
1690                      "0x" RAM_ADDR_FMT " or maxmem size "
1691                      "0x" RAM_ADDR_FMT " isn't aligned to %llu MB",
1692                      machine->ram_size, machine->maxram_size,
1693                      SPAPR_MEMORY_BLOCK_SIZE/M_BYTE);
1694         exit(EXIT_FAILURE);
1695     }
1696 
1697     for (i = 0; i < nb_numa_nodes; i++) {
1698         if (numa_info[i].node_mem % SPAPR_MEMORY_BLOCK_SIZE) {
1699             error_report("Can't support memory configuration where memory size"
1700                          " %" PRIx64 " of node %d isn't aligned to %llu MB",
1701                          numa_info[i].node_mem, i,
1702                          SPAPR_MEMORY_BLOCK_SIZE/M_BYTE);
1703             exit(EXIT_FAILURE);
1704         }
1705     }
1706 }
1707 
1708 /* pSeries LPAR / sPAPR hardware init */
1709 static void ppc_spapr_init(MachineState *machine)
1710 {
1711     sPAPRMachineState *spapr = SPAPR_MACHINE(machine);
1712     sPAPRMachineClass *smc = SPAPR_MACHINE_GET_CLASS(machine);
1713     const char *kernel_filename = machine->kernel_filename;
1714     const char *kernel_cmdline = machine->kernel_cmdline;
1715     const char *initrd_filename = machine->initrd_filename;
1716     PowerPCCPU *cpu;
1717     PCIHostState *phb;
1718     int i;
1719     MemoryRegion *sysmem = get_system_memory();
1720     MemoryRegion *ram = g_new(MemoryRegion, 1);
1721     MemoryRegion *rma_region;
1722     void *rma = NULL;
1723     hwaddr rma_alloc_size;
1724     hwaddr node0_size = spapr_node0_size();
1725     uint32_t initrd_base = 0;
1726     long kernel_size = 0, initrd_size = 0;
1727     long load_limit, fw_size;
1728     bool kernel_le = false;
1729     char *filename;
1730 
1731     msi_supported = true;
1732 
1733     QLIST_INIT(&spapr->phbs);
1734 
1735     cpu_ppc_hypercall = emulate_spapr_hypercall;
1736 
1737     /* Allocate RMA if necessary */
1738     rma_alloc_size = kvmppc_alloc_rma(&rma);
1739 
1740     if (rma_alloc_size == -1) {
1741         error_report("Unable to create RMA");
1742         exit(1);
1743     }
1744 
1745     if (rma_alloc_size && (rma_alloc_size < node0_size)) {
1746         spapr->rma_size = rma_alloc_size;
1747     } else {
1748         spapr->rma_size = node0_size;
1749 
1750         /* With KVM, we don't actually know whether KVM supports an
1751          * unbounded RMA (PR KVM) or is limited by the hash table size
1752          * (HV KVM using VRMA), so we always assume the latter
1753          *
1754          * In that case, we also limit the initial allocations for RTAS
1755          * etc... to 256M since we have no way to know what the VRMA size
1756          * is going to be as it depends on the size of the hash table
1757          * isn't determined yet.
1758          */
1759         if (kvm_enabled()) {
1760             spapr->vrma_adjust = 1;
1761             spapr->rma_size = MIN(spapr->rma_size, 0x10000000);
1762         }
1763     }
1764 
1765     if (spapr->rma_size > node0_size) {
1766         fprintf(stderr, "Error: Numa node 0 has to span the RMA (%#08"HWADDR_PRIx")\n",
1767                 spapr->rma_size);
1768         exit(1);
1769     }
1770 
1771     /* Setup a load limit for the ramdisk leaving room for SLOF and FDT */
1772     load_limit = MIN(spapr->rma_size, RTAS_MAX_ADDR) - FW_OVERHEAD;
1773 
1774     /* We aim for a hash table of size 1/128 the size of RAM.  The
1775      * normal rule of thumb is 1/64 the size of RAM, but that's much
1776      * more than needed for the Linux guests we support. */
1777     spapr->htab_shift = 18; /* Minimum architected size */
1778     while (spapr->htab_shift <= 46) {
1779         if ((1ULL << (spapr->htab_shift + 7)) >= machine->maxram_size) {
1780             break;
1781         }
1782         spapr->htab_shift++;
1783     }
1784     spapr_alloc_htab(spapr);
1785 
1786     /* Set up Interrupt Controller before we create the VCPUs */
1787     spapr->icp = xics_system_init(machine,
1788                                   DIV_ROUND_UP(max_cpus * kvmppc_smt_threads(),
1789                                                smp_threads),
1790                                   XICS_IRQS);
1791 
1792     if (smc->dr_lmb_enabled) {
1793         spapr_validate_node_memory(machine);
1794     }
1795 
1796     /* init CPUs */
1797     if (machine->cpu_model == NULL) {
1798         machine->cpu_model = kvm_enabled() ? "host" : "POWER7";
1799     }
1800     for (i = 0; i < smp_cpus; i++) {
1801         cpu = cpu_ppc_init(machine->cpu_model);
1802         if (cpu == NULL) {
1803             fprintf(stderr, "Unable to find PowerPC CPU definition\n");
1804             exit(1);
1805         }
1806         spapr_cpu_init(spapr, cpu);
1807     }
1808 
1809     if (kvm_enabled()) {
1810         /* Enable H_LOGICAL_CI_* so SLOF can talk to in-kernel devices */
1811         kvmppc_enable_logical_ci_hcalls();
1812         kvmppc_enable_set_mode_hcall();
1813     }
1814 
1815     /* allocate RAM */
1816     memory_region_allocate_system_memory(ram, NULL, "ppc_spapr.ram",
1817                                          machine->ram_size);
1818     memory_region_add_subregion(sysmem, 0, ram);
1819 
1820     if (rma_alloc_size && rma) {
1821         rma_region = g_new(MemoryRegion, 1);
1822         memory_region_init_ram_ptr(rma_region, NULL, "ppc_spapr.rma",
1823                                    rma_alloc_size, rma);
1824         vmstate_register_ram_global(rma_region);
1825         memory_region_add_subregion(sysmem, 0, rma_region);
1826     }
1827 
1828     /* initialize hotplug memory address space */
1829     if (machine->ram_size < machine->maxram_size) {
1830         ram_addr_t hotplug_mem_size = machine->maxram_size - machine->ram_size;
1831 
1832         if (machine->ram_slots > SPAPR_MAX_RAM_SLOTS) {
1833             error_report("Specified number of memory slots %"PRIu64" exceeds max supported %d\n",
1834                          machine->ram_slots, SPAPR_MAX_RAM_SLOTS);
1835             exit(EXIT_FAILURE);
1836         }
1837 
1838         spapr->hotplug_memory.base = ROUND_UP(machine->ram_size,
1839                                               SPAPR_HOTPLUG_MEM_ALIGN);
1840         memory_region_init(&spapr->hotplug_memory.mr, OBJECT(spapr),
1841                            "hotplug-memory", hotplug_mem_size);
1842         memory_region_add_subregion(sysmem, spapr->hotplug_memory.base,
1843                                     &spapr->hotplug_memory.mr);
1844     }
1845 
1846     if (smc->dr_lmb_enabled) {
1847         spapr_create_lmb_dr_connectors(spapr);
1848     }
1849 
1850     filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, "spapr-rtas.bin");
1851     if (!filename) {
1852         error_report("Could not find LPAR rtas '%s'", "spapr-rtas.bin");
1853         exit(1);
1854     }
1855     spapr->rtas_size = get_image_size(filename);
1856     spapr->rtas_blob = g_malloc(spapr->rtas_size);
1857     if (load_image_size(filename, spapr->rtas_blob, spapr->rtas_size) < 0) {
1858         error_report("Could not load LPAR rtas '%s'", filename);
1859         exit(1);
1860     }
1861     if (spapr->rtas_size > RTAS_MAX_SIZE) {
1862         error_report("RTAS too big ! 0x%zx bytes (max is 0x%x)",
1863                      (size_t)spapr->rtas_size, RTAS_MAX_SIZE);
1864         exit(1);
1865     }
1866     g_free(filename);
1867 
1868     /* Set up EPOW events infrastructure */
1869     spapr_events_init(spapr);
1870 
1871     /* Set up the RTC RTAS interfaces */
1872     spapr_rtc_create(spapr);
1873 
1874     /* Set up VIO bus */
1875     spapr->vio_bus = spapr_vio_bus_init();
1876 
1877     for (i = 0; i < MAX_SERIAL_PORTS; i++) {
1878         if (serial_hds[i]) {
1879             spapr_vty_create(spapr->vio_bus, serial_hds[i]);
1880         }
1881     }
1882 
1883     /* We always have at least the nvram device on VIO */
1884     spapr_create_nvram(spapr);
1885 
1886     /* Set up PCI */
1887     spapr_pci_rtas_init();
1888 
1889     phb = spapr_create_phb(spapr, 0);
1890 
1891     for (i = 0; i < nb_nics; i++) {
1892         NICInfo *nd = &nd_table[i];
1893 
1894         if (!nd->model) {
1895             nd->model = g_strdup("ibmveth");
1896         }
1897 
1898         if (strcmp(nd->model, "ibmveth") == 0) {
1899             spapr_vlan_create(spapr->vio_bus, nd);
1900         } else {
1901             pci_nic_init_nofail(&nd_table[i], phb->bus, nd->model, NULL);
1902         }
1903     }
1904 
1905     for (i = 0; i <= drive_get_max_bus(IF_SCSI); i++) {
1906         spapr_vscsi_create(spapr->vio_bus);
1907     }
1908 
1909     /* Graphics */
1910     if (spapr_vga_init(phb->bus)) {
1911         spapr->has_graphics = true;
1912         machine->usb |= defaults_enabled() && !machine->usb_disabled;
1913     }
1914 
1915     if (machine->usb) {
1916         pci_create_simple(phb->bus, -1, "pci-ohci");
1917 
1918         if (spapr->has_graphics) {
1919             USBBus *usb_bus = usb_bus_find(-1);
1920 
1921             usb_create_simple(usb_bus, "usb-kbd");
1922             usb_create_simple(usb_bus, "usb-mouse");
1923         }
1924     }
1925 
1926     if (spapr->rma_size < (MIN_RMA_SLOF << 20)) {
1927         fprintf(stderr, "qemu: pSeries SLOF firmware requires >= "
1928                 "%ldM guest RMA (Real Mode Area memory)\n", MIN_RMA_SLOF);
1929         exit(1);
1930     }
1931 
1932     if (kernel_filename) {
1933         uint64_t lowaddr = 0;
1934 
1935         kernel_size = load_elf(kernel_filename, translate_kernel_address, NULL,
1936                                NULL, &lowaddr, NULL, 1, PPC_ELF_MACHINE, 0);
1937         if (kernel_size == ELF_LOAD_WRONG_ENDIAN) {
1938             kernel_size = load_elf(kernel_filename,
1939                                    translate_kernel_address, NULL,
1940                                    NULL, &lowaddr, NULL, 0, PPC_ELF_MACHINE, 0);
1941             kernel_le = kernel_size > 0;
1942         }
1943         if (kernel_size < 0) {
1944             fprintf(stderr, "qemu: error loading %s: %s\n",
1945                     kernel_filename, load_elf_strerror(kernel_size));
1946             exit(1);
1947         }
1948 
1949         /* load initrd */
1950         if (initrd_filename) {
1951             /* Try to locate the initrd in the gap between the kernel
1952              * and the firmware. Add a bit of space just in case
1953              */
1954             initrd_base = (KERNEL_LOAD_ADDR + kernel_size + 0x1ffff) & ~0xffff;
1955             initrd_size = load_image_targphys(initrd_filename, initrd_base,
1956                                               load_limit - initrd_base);
1957             if (initrd_size < 0) {
1958                 fprintf(stderr, "qemu: could not load initial ram disk '%s'\n",
1959                         initrd_filename);
1960                 exit(1);
1961             }
1962         } else {
1963             initrd_base = 0;
1964             initrd_size = 0;
1965         }
1966     }
1967 
1968     if (bios_name == NULL) {
1969         bios_name = FW_FILE_NAME;
1970     }
1971     filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, bios_name);
1972     if (!filename) {
1973         error_report("Could not find LPAR firmware '%s'", bios_name);
1974         exit(1);
1975     }
1976     fw_size = load_image_targphys(filename, 0, FW_MAX_SIZE);
1977     if (fw_size <= 0) {
1978         error_report("Could not load LPAR firmware '%s'", filename);
1979         exit(1);
1980     }
1981     g_free(filename);
1982 
1983     /* FIXME: Should register things through the MachineState's qdev
1984      * interface, this is a legacy from the sPAPREnvironment structure
1985      * which predated MachineState but had a similar function */
1986     vmstate_register(NULL, 0, &vmstate_spapr, spapr);
1987     register_savevm_live(NULL, "spapr/htab", -1, 1,
1988                          &savevm_htab_handlers, spapr);
1989 
1990     /* Prepare the device tree */
1991     spapr->fdt_skel = spapr_create_fdt_skel(initrd_base, initrd_size,
1992                                             kernel_size, kernel_le,
1993                                             kernel_cmdline,
1994                                             spapr->check_exception_irq);
1995     assert(spapr->fdt_skel != NULL);
1996 
1997     /* used by RTAS */
1998     QTAILQ_INIT(&spapr->ccs_list);
1999     qemu_register_reset(spapr_ccs_reset_hook, spapr);
2000 
2001     qemu_register_boot_set(spapr_boot_set, spapr);
2002 }
2003 
2004 static int spapr_kvm_type(const char *vm_type)
2005 {
2006     if (!vm_type) {
2007         return 0;
2008     }
2009 
2010     if (!strcmp(vm_type, "HV")) {
2011         return 1;
2012     }
2013 
2014     if (!strcmp(vm_type, "PR")) {
2015         return 2;
2016     }
2017 
2018     error_report("Unknown kvm-type specified '%s'", vm_type);
2019     exit(1);
2020 }
2021 
2022 /*
2023  * Implementation of an interface to adjust firmware path
2024  * for the bootindex property handling.
2025  */
2026 static char *spapr_get_fw_dev_path(FWPathProvider *p, BusState *bus,
2027                                    DeviceState *dev)
2028 {
2029 #define CAST(type, obj, name) \
2030     ((type *)object_dynamic_cast(OBJECT(obj), (name)))
2031     SCSIDevice *d = CAST(SCSIDevice,  dev, TYPE_SCSI_DEVICE);
2032     sPAPRPHBState *phb = CAST(sPAPRPHBState, dev, TYPE_SPAPR_PCI_HOST_BRIDGE);
2033 
2034     if (d) {
2035         void *spapr = CAST(void, bus->parent, "spapr-vscsi");
2036         VirtIOSCSI *virtio = CAST(VirtIOSCSI, bus->parent, TYPE_VIRTIO_SCSI);
2037         USBDevice *usb = CAST(USBDevice, bus->parent, TYPE_USB_DEVICE);
2038 
2039         if (spapr) {
2040             /*
2041              * Replace "channel@0/disk@0,0" with "disk@8000000000000000":
2042              * We use SRP luns of the form 8000 | (bus << 8) | (id << 5) | lun
2043              * in the top 16 bits of the 64-bit LUN
2044              */
2045             unsigned id = 0x8000 | (d->id << 8) | d->lun;
2046             return g_strdup_printf("%s@%"PRIX64, qdev_fw_name(dev),
2047                                    (uint64_t)id << 48);
2048         } else if (virtio) {
2049             /*
2050              * We use SRP luns of the form 01000000 | (target << 8) | lun
2051              * in the top 32 bits of the 64-bit LUN
2052              * Note: the quote above is from SLOF and it is wrong,
2053              * the actual binding is:
2054              * swap 0100 or 10 << or 20 << ( target lun-id -- srplun )
2055              */
2056             unsigned id = 0x1000000 | (d->id << 16) | d->lun;
2057             return g_strdup_printf("%s@%"PRIX64, qdev_fw_name(dev),
2058                                    (uint64_t)id << 32);
2059         } else if (usb) {
2060             /*
2061              * We use SRP luns of the form 01000000 | (usb-port << 16) | lun
2062              * in the top 32 bits of the 64-bit LUN
2063              */
2064             unsigned usb_port = atoi(usb->port->path);
2065             unsigned id = 0x1000000 | (usb_port << 16) | d->lun;
2066             return g_strdup_printf("%s@%"PRIX64, qdev_fw_name(dev),
2067                                    (uint64_t)id << 32);
2068         }
2069     }
2070 
2071     if (phb) {
2072         /* Replace "pci" with "pci@800000020000000" */
2073         return g_strdup_printf("pci@%"PRIX64, phb->buid);
2074     }
2075 
2076     return NULL;
2077 }
2078 
2079 static char *spapr_get_kvm_type(Object *obj, Error **errp)
2080 {
2081     sPAPRMachineState *spapr = SPAPR_MACHINE(obj);
2082 
2083     return g_strdup(spapr->kvm_type);
2084 }
2085 
2086 static void spapr_set_kvm_type(Object *obj, const char *value, Error **errp)
2087 {
2088     sPAPRMachineState *spapr = SPAPR_MACHINE(obj);
2089 
2090     g_free(spapr->kvm_type);
2091     spapr->kvm_type = g_strdup(value);
2092 }
2093 
2094 static void spapr_machine_initfn(Object *obj)
2095 {
2096     object_property_add_str(obj, "kvm-type",
2097                             spapr_get_kvm_type, spapr_set_kvm_type, NULL);
2098     object_property_set_description(obj, "kvm-type",
2099                                     "Specifies the KVM virtualization mode (HV, PR)",
2100                                     NULL);
2101 }
2102 
2103 static void ppc_cpu_do_nmi_on_cpu(void *arg)
2104 {
2105     CPUState *cs = arg;
2106 
2107     cpu_synchronize_state(cs);
2108     ppc_cpu_do_system_reset(cs);
2109 }
2110 
2111 static void spapr_nmi(NMIState *n, int cpu_index, Error **errp)
2112 {
2113     CPUState *cs;
2114 
2115     CPU_FOREACH(cs) {
2116         async_run_on_cpu(cs, ppc_cpu_do_nmi_on_cpu, cs);
2117     }
2118 }
2119 
2120 static void spapr_add_lmbs(DeviceState *dev, uint64_t addr, uint64_t size,
2121                            uint32_t node, Error **errp)
2122 {
2123     sPAPRDRConnector *drc;
2124     sPAPRDRConnectorClass *drck;
2125     uint32_t nr_lmbs = size/SPAPR_MEMORY_BLOCK_SIZE;
2126     int i, fdt_offset, fdt_size;
2127     void *fdt;
2128 
2129     /*
2130      * Check for DRC connectors and send hotplug notification to the
2131      * guest only in case of hotplugged memory. This allows cold plugged
2132      * memory to be specified at boot time.
2133      */
2134     if (!dev->hotplugged) {
2135         return;
2136     }
2137 
2138     for (i = 0; i < nr_lmbs; i++) {
2139         drc = spapr_dr_connector_by_id(SPAPR_DR_CONNECTOR_TYPE_LMB,
2140                 addr/SPAPR_MEMORY_BLOCK_SIZE);
2141         g_assert(drc);
2142 
2143         fdt = create_device_tree(&fdt_size);
2144         fdt_offset = spapr_populate_memory_node(fdt, node, addr,
2145                                                 SPAPR_MEMORY_BLOCK_SIZE);
2146 
2147         drck = SPAPR_DR_CONNECTOR_GET_CLASS(drc);
2148         drck->attach(drc, dev, fdt, fdt_offset, !dev->hotplugged, errp);
2149         addr += SPAPR_MEMORY_BLOCK_SIZE;
2150     }
2151     spapr_hotplug_req_add_by_count(SPAPR_DR_CONNECTOR_TYPE_LMB, nr_lmbs);
2152 }
2153 
2154 static void spapr_memory_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
2155                               uint32_t node, Error **errp)
2156 {
2157     Error *local_err = NULL;
2158     sPAPRMachineState *ms = SPAPR_MACHINE(hotplug_dev);
2159     PCDIMMDevice *dimm = PC_DIMM(dev);
2160     PCDIMMDeviceClass *ddc = PC_DIMM_GET_CLASS(dimm);
2161     MemoryRegion *mr = ddc->get_memory_region(dimm);
2162     uint64_t align = memory_region_get_alignment(mr);
2163     uint64_t size = memory_region_size(mr);
2164     uint64_t addr;
2165 
2166     if (size % SPAPR_MEMORY_BLOCK_SIZE) {
2167         error_setg(&local_err, "Hotplugged memory size must be a multiple of "
2168                       "%lld MB", SPAPR_MEMORY_BLOCK_SIZE/M_BYTE);
2169         goto out;
2170     }
2171 
2172     pc_dimm_memory_plug(dev, &ms->hotplug_memory, mr, align, &local_err);
2173     if (local_err) {
2174         goto out;
2175     }
2176 
2177     addr = object_property_get_int(OBJECT(dimm), PC_DIMM_ADDR_PROP, &local_err);
2178     if (local_err) {
2179         pc_dimm_memory_unplug(dev, &ms->hotplug_memory, mr);
2180         goto out;
2181     }
2182 
2183     spapr_add_lmbs(dev, addr, size, node, &error_abort);
2184 
2185 out:
2186     error_propagate(errp, local_err);
2187 }
2188 
2189 static void spapr_machine_device_plug(HotplugHandler *hotplug_dev,
2190                                       DeviceState *dev, Error **errp)
2191 {
2192     sPAPRMachineClass *smc = SPAPR_MACHINE_GET_CLASS(qdev_get_machine());
2193 
2194     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2195         int node;
2196 
2197         if (!smc->dr_lmb_enabled) {
2198             error_setg(errp, "Memory hotplug not supported for this machine");
2199             return;
2200         }
2201         node = object_property_get_int(OBJECT(dev), PC_DIMM_NODE_PROP, errp);
2202         if (*errp) {
2203             return;
2204         }
2205 
2206         /*
2207          * Currently PowerPC kernel doesn't allow hot-adding memory to
2208          * memory-less node, but instead will silently add the memory
2209          * to the first node that has some memory. This causes two
2210          * unexpected behaviours for the user.
2211          *
2212          * - Memory gets hotplugged to a different node than what the user
2213          *   specified.
2214          * - Since pc-dimm subsystem in QEMU still thinks that memory belongs
2215          *   to memory-less node, a reboot will set things accordingly
2216          *   and the previously hotplugged memory now ends in the right node.
2217          *   This appears as if some memory moved from one node to another.
2218          *
2219          * So until kernel starts supporting memory hotplug to memory-less
2220          * nodes, just prevent such attempts upfront in QEMU.
2221          */
2222         if (nb_numa_nodes && !numa_info[node].node_mem) {
2223             error_setg(errp, "Can't hotplug memory to memory-less node %d",
2224                        node);
2225             return;
2226         }
2227 
2228         spapr_memory_plug(hotplug_dev, dev, node, errp);
2229     }
2230 }
2231 
2232 static void spapr_machine_device_unplug(HotplugHandler *hotplug_dev,
2233                                       DeviceState *dev, Error **errp)
2234 {
2235     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2236         error_setg(errp, "Memory hot unplug not supported by sPAPR");
2237     }
2238 }
2239 
2240 static HotplugHandler *spapr_get_hotpug_handler(MachineState *machine,
2241                                              DeviceState *dev)
2242 {
2243     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2244         return HOTPLUG_HANDLER(machine);
2245     }
2246     return NULL;
2247 }
2248 
2249 static unsigned spapr_cpu_index_to_socket_id(unsigned cpu_index)
2250 {
2251     /* Allocate to NUMA nodes on a "socket" basis (not that concept of
2252      * socket means much for the paravirtualized PAPR platform) */
2253     return cpu_index / smp_threads / smp_cores;
2254 }
2255 
2256 static void spapr_machine_class_init(ObjectClass *oc, void *data)
2257 {
2258     MachineClass *mc = MACHINE_CLASS(oc);
2259     sPAPRMachineClass *smc = SPAPR_MACHINE_CLASS(oc);
2260     FWPathProviderClass *fwc = FW_PATH_PROVIDER_CLASS(oc);
2261     NMIClass *nc = NMI_CLASS(oc);
2262     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
2263 
2264     mc->init = ppc_spapr_init;
2265     mc->reset = ppc_spapr_reset;
2266     mc->block_default_type = IF_SCSI;
2267     mc->max_cpus = MAX_CPUMASK_BITS;
2268     mc->no_parallel = 1;
2269     mc->default_boot_order = "";
2270     mc->default_ram_size = 512 * M_BYTE;
2271     mc->kvm_type = spapr_kvm_type;
2272     mc->has_dynamic_sysbus = true;
2273     mc->pci_allow_0_address = true;
2274     mc->get_hotplug_handler = spapr_get_hotpug_handler;
2275     hc->plug = spapr_machine_device_plug;
2276     hc->unplug = spapr_machine_device_unplug;
2277     mc->cpu_index_to_socket_id = spapr_cpu_index_to_socket_id;
2278 
2279     smc->dr_lmb_enabled = false;
2280     fwc->get_dev_path = spapr_get_fw_dev_path;
2281     nc->nmi_monitor_handler = spapr_nmi;
2282 }
2283 
2284 static const TypeInfo spapr_machine_info = {
2285     .name          = TYPE_SPAPR_MACHINE,
2286     .parent        = TYPE_MACHINE,
2287     .abstract      = true,
2288     .instance_size = sizeof(sPAPRMachineState),
2289     .instance_init = spapr_machine_initfn,
2290     .class_size    = sizeof(sPAPRMachineClass),
2291     .class_init    = spapr_machine_class_init,
2292     .interfaces = (InterfaceInfo[]) {
2293         { TYPE_FW_PATH_PROVIDER },
2294         { TYPE_NMI },
2295         { TYPE_HOTPLUG_HANDLER },
2296         { }
2297     },
2298 };
2299 
2300 #define SPAPR_COMPAT_2_4 \
2301         HW_COMPAT_2_4
2302 
2303 #define SPAPR_COMPAT_2_3 \
2304         SPAPR_COMPAT_2_4 \
2305         HW_COMPAT_2_3 \
2306         {\
2307             .driver   = "spapr-pci-host-bridge",\
2308             .property = "dynamic-reconfiguration",\
2309             .value    = "off",\
2310         },
2311 
2312 #define SPAPR_COMPAT_2_2 \
2313         SPAPR_COMPAT_2_3 \
2314         HW_COMPAT_2_2 \
2315         {\
2316             .driver   = TYPE_SPAPR_PCI_HOST_BRIDGE,\
2317             .property = "mem_win_size",\
2318             .value    = "0x20000000",\
2319         },
2320 
2321 #define SPAPR_COMPAT_2_1 \
2322         SPAPR_COMPAT_2_2 \
2323         HW_COMPAT_2_1
2324 
2325 static void spapr_compat_2_3(Object *obj)
2326 {
2327     savevm_skip_section_footers();
2328     global_state_set_optional();
2329 }
2330 
2331 static void spapr_compat_2_2(Object *obj)
2332 {
2333     spapr_compat_2_3(obj);
2334 }
2335 
2336 static void spapr_compat_2_1(Object *obj)
2337 {
2338     spapr_compat_2_2(obj);
2339 }
2340 
2341 static void spapr_machine_2_3_instance_init(Object *obj)
2342 {
2343     spapr_compat_2_3(obj);
2344     spapr_machine_initfn(obj);
2345 }
2346 
2347 static void spapr_machine_2_2_instance_init(Object *obj)
2348 {
2349     spapr_compat_2_2(obj);
2350     spapr_machine_initfn(obj);
2351 }
2352 
2353 static void spapr_machine_2_1_instance_init(Object *obj)
2354 {
2355     spapr_compat_2_1(obj);
2356     spapr_machine_initfn(obj);
2357 }
2358 
2359 static void spapr_machine_2_1_class_init(ObjectClass *oc, void *data)
2360 {
2361     MachineClass *mc = MACHINE_CLASS(oc);
2362     static GlobalProperty compat_props[] = {
2363         SPAPR_COMPAT_2_1
2364         { /* end of list */ }
2365     };
2366 
2367     mc->desc = "pSeries Logical Partition (PAPR compliant) v2.1";
2368     mc->compat_props = compat_props;
2369 }
2370 
2371 static const TypeInfo spapr_machine_2_1_info = {
2372     .name          = MACHINE_TYPE_NAME("pseries-2.1"),
2373     .parent        = TYPE_SPAPR_MACHINE,
2374     .class_init    = spapr_machine_2_1_class_init,
2375     .instance_init = spapr_machine_2_1_instance_init,
2376 };
2377 
2378 static void spapr_machine_2_2_class_init(ObjectClass *oc, void *data)
2379 {
2380     static GlobalProperty compat_props[] = {
2381         SPAPR_COMPAT_2_2
2382         { /* end of list */ }
2383     };
2384     MachineClass *mc = MACHINE_CLASS(oc);
2385 
2386     mc->desc = "pSeries Logical Partition (PAPR compliant) v2.2";
2387     mc->compat_props = compat_props;
2388 }
2389 
2390 static const TypeInfo spapr_machine_2_2_info = {
2391     .name          = MACHINE_TYPE_NAME("pseries-2.2"),
2392     .parent        = TYPE_SPAPR_MACHINE,
2393     .class_init    = spapr_machine_2_2_class_init,
2394     .instance_init = spapr_machine_2_2_instance_init,
2395 };
2396 
2397 static void spapr_machine_2_3_class_init(ObjectClass *oc, void *data)
2398 {
2399     static GlobalProperty compat_props[] = {
2400         SPAPR_COMPAT_2_3
2401         { /* end of list */ }
2402     };
2403     MachineClass *mc = MACHINE_CLASS(oc);
2404 
2405     mc->desc = "pSeries Logical Partition (PAPR compliant) v2.3";
2406     mc->compat_props = compat_props;
2407 }
2408 
2409 static const TypeInfo spapr_machine_2_3_info = {
2410     .name          = MACHINE_TYPE_NAME("pseries-2.3"),
2411     .parent        = TYPE_SPAPR_MACHINE,
2412     .class_init    = spapr_machine_2_3_class_init,
2413     .instance_init = spapr_machine_2_3_instance_init,
2414 };
2415 
2416 static void spapr_machine_2_4_class_init(ObjectClass *oc, void *data)
2417 {
2418     static GlobalProperty compat_props[] = {
2419         SPAPR_COMPAT_2_4
2420         { /* end of list */ }
2421     };
2422     MachineClass *mc = MACHINE_CLASS(oc);
2423 
2424     mc->desc = "pSeries Logical Partition (PAPR compliant) v2.4";
2425     mc->alias = "pseries";
2426     mc->is_default = 0;
2427     mc->compat_props = compat_props;
2428 }
2429 
2430 static const TypeInfo spapr_machine_2_4_info = {
2431     .name          = MACHINE_TYPE_NAME("pseries-2.4"),
2432     .parent        = TYPE_SPAPR_MACHINE,
2433     .class_init    = spapr_machine_2_4_class_init,
2434 };
2435 
2436 static void spapr_machine_2_5_class_init(ObjectClass *oc, void *data)
2437 {
2438     MachineClass *mc = MACHINE_CLASS(oc);
2439     sPAPRMachineClass *smc = SPAPR_MACHINE_CLASS(oc);
2440 
2441     mc->name = "pseries-2.5";
2442     mc->desc = "pSeries Logical Partition (PAPR compliant) v2.5";
2443     mc->alias = "pseries";
2444     mc->is_default = 1;
2445     smc->dr_lmb_enabled = true;
2446 }
2447 
2448 static const TypeInfo spapr_machine_2_5_info = {
2449     .name          = MACHINE_TYPE_NAME("pseries-2.5"),
2450     .parent        = TYPE_SPAPR_MACHINE,
2451     .class_init    = spapr_machine_2_5_class_init,
2452 };
2453 
2454 static void spapr_machine_register_types(void)
2455 {
2456     type_register_static(&spapr_machine_info);
2457     type_register_static(&spapr_machine_2_1_info);
2458     type_register_static(&spapr_machine_2_2_info);
2459     type_register_static(&spapr_machine_2_3_info);
2460     type_register_static(&spapr_machine_2_4_info);
2461     type_register_static(&spapr_machine_2_5_info);
2462 }
2463 
2464 type_init(spapr_machine_register_types)
2465