xref: /qemu/hw/ppc/spapr.c (revision b1be0972)
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 "qemu/osdep.h"
28 #include "sysemu/sysemu.h"
29 #include "sysemu/numa.h"
30 #include "hw/hw.h"
31 #include "hw/fw-path-provider.h"
32 #include "elf.h"
33 #include "net/net.h"
34 #include "sysemu/device_tree.h"
35 #include "sysemu/block-backend.h"
36 #include "sysemu/cpus.h"
37 #include "sysemu/kvm.h"
38 #include "sysemu/device_tree.h"
39 #include "kvm_ppc.h"
40 #include "migration/migration.h"
41 #include "mmu-hash64.h"
42 #include "qom/cpu.h"
43 
44 #include "hw/boards.h"
45 #include "hw/ppc/ppc.h"
46 #include "hw/loader.h"
47 
48 #include "hw/ppc/spapr.h"
49 #include "hw/ppc/spapr_vio.h"
50 #include "hw/pci-host/spapr.h"
51 #include "hw/ppc/xics.h"
52 #include "hw/pci/msi.h"
53 
54 #include "hw/pci/pci.h"
55 #include "hw/scsi/scsi.h"
56 #include "hw/virtio/virtio-scsi.h"
57 
58 #include "exec/address-spaces.h"
59 #include "hw/usb.h"
60 #include "qemu/config-file.h"
61 #include "qemu/error-report.h"
62 #include "trace.h"
63 #include "hw/nmi.h"
64 
65 #include "hw/compat.h"
66 #include "qemu-common.h"
67 
68 #include <libfdt.h>
69 
70 /* SLOF memory layout:
71  *
72  * SLOF raw image loaded at 0, copies its romfs right below the flat
73  * device-tree, then position SLOF itself 31M below that
74  *
75  * So we set FW_OVERHEAD to 40MB which should account for all of that
76  * and more
77  *
78  * We load our kernel at 4M, leaving space for SLOF initial image
79  */
80 #define FDT_MAX_SIZE            0x100000
81 #define RTAS_MAX_SIZE           0x10000
82 #define RTAS_MAX_ADDR           0x80000000 /* RTAS must stay below that */
83 #define FW_MAX_SIZE             0x400000
84 #define FW_FILE_NAME            "slof.bin"
85 #define FW_OVERHEAD             0x2800000
86 #define KERNEL_LOAD_ADDR        FW_MAX_SIZE
87 
88 #define MIN_RMA_SLOF            128UL
89 
90 #define TIMEBASE_FREQ           512000000ULL
91 
92 #define PHANDLE_XICP            0x00001111
93 
94 #define HTAB_SIZE(spapr)        (1ULL << ((spapr)->htab_shift))
95 
96 static XICSState *try_create_xics(const char *type, int nr_servers,
97                                   int nr_irqs, Error **errp)
98 {
99     Error *err = NULL;
100     DeviceState *dev;
101 
102     dev = qdev_create(NULL, type);
103     qdev_prop_set_uint32(dev, "nr_servers", nr_servers);
104     qdev_prop_set_uint32(dev, "nr_irqs", nr_irqs);
105     object_property_set_bool(OBJECT(dev), true, "realized", &err);
106     if (err) {
107         error_propagate(errp, err);
108         object_unparent(OBJECT(dev));
109         return NULL;
110     }
111     return XICS_COMMON(dev);
112 }
113 
114 static XICSState *xics_system_init(MachineState *machine,
115                                    int nr_servers, int nr_irqs, Error **errp)
116 {
117     XICSState *icp = NULL;
118 
119     if (kvm_enabled()) {
120         Error *err = NULL;
121 
122         if (machine_kernel_irqchip_allowed(machine)) {
123             icp = try_create_xics(TYPE_KVM_XICS, nr_servers, nr_irqs, &err);
124         }
125         if (machine_kernel_irqchip_required(machine) && !icp) {
126             error_reportf_err(err,
127                               "kernel_irqchip requested but unavailable: ");
128         } else {
129             error_free(err);
130         }
131     }
132 
133     if (!icp) {
134         icp = try_create_xics(TYPE_XICS, nr_servers, nr_irqs, errp);
135     }
136 
137     return icp;
138 }
139 
140 static int spapr_fixup_cpu_smt_dt(void *fdt, int offset, PowerPCCPU *cpu,
141                                   int smt_threads)
142 {
143     int i, ret = 0;
144     uint32_t servers_prop[smt_threads];
145     uint32_t gservers_prop[smt_threads * 2];
146     int index = ppc_get_vcpu_dt_id(cpu);
147 
148     if (cpu->cpu_version) {
149         ret = fdt_setprop_cell(fdt, offset, "cpu-version", cpu->cpu_version);
150         if (ret < 0) {
151             return ret;
152         }
153     }
154 
155     /* Build interrupt servers and gservers properties */
156     for (i = 0; i < smt_threads; i++) {
157         servers_prop[i] = cpu_to_be32(index + i);
158         /* Hack, direct the group queues back to cpu 0 */
159         gservers_prop[i*2] = cpu_to_be32(index + i);
160         gservers_prop[i*2 + 1] = 0;
161     }
162     ret = fdt_setprop(fdt, offset, "ibm,ppc-interrupt-server#s",
163                       servers_prop, sizeof(servers_prop));
164     if (ret < 0) {
165         return ret;
166     }
167     ret = fdt_setprop(fdt, offset, "ibm,ppc-interrupt-gserver#s",
168                       gservers_prop, sizeof(gservers_prop));
169 
170     return ret;
171 }
172 
173 static int spapr_fixup_cpu_numa_dt(void *fdt, int offset, CPUState *cs)
174 {
175     int ret = 0;
176     PowerPCCPU *cpu = POWERPC_CPU(cs);
177     int index = ppc_get_vcpu_dt_id(cpu);
178     uint32_t associativity[] = {cpu_to_be32(0x5),
179                                 cpu_to_be32(0x0),
180                                 cpu_to_be32(0x0),
181                                 cpu_to_be32(0x0),
182                                 cpu_to_be32(cs->numa_node),
183                                 cpu_to_be32(index)};
184 
185     /* Advertise NUMA via ibm,associativity */
186     if (nb_numa_nodes > 1) {
187         ret = fdt_setprop(fdt, offset, "ibm,associativity", associativity,
188                           sizeof(associativity));
189     }
190 
191     return ret;
192 }
193 
194 static int spapr_fixup_cpu_dt(void *fdt, sPAPRMachineState *spapr)
195 {
196     int ret = 0, offset, cpus_offset;
197     CPUState *cs;
198     char cpu_model[32];
199     int smt = kvmppc_smt_threads();
200     uint32_t pft_size_prop[] = {0, cpu_to_be32(spapr->htab_shift)};
201 
202     CPU_FOREACH(cs) {
203         PowerPCCPU *cpu = POWERPC_CPU(cs);
204         DeviceClass *dc = DEVICE_GET_CLASS(cs);
205         int index = ppc_get_vcpu_dt_id(cpu);
206 
207         if ((index % smt) != 0) {
208             continue;
209         }
210 
211         snprintf(cpu_model, 32, "%s@%x", dc->fw_name, index);
212 
213         cpus_offset = fdt_path_offset(fdt, "/cpus");
214         if (cpus_offset < 0) {
215             cpus_offset = fdt_add_subnode(fdt, fdt_path_offset(fdt, "/"),
216                                           "cpus");
217             if (cpus_offset < 0) {
218                 return cpus_offset;
219             }
220         }
221         offset = fdt_subnode_offset(fdt, cpus_offset, cpu_model);
222         if (offset < 0) {
223             offset = fdt_add_subnode(fdt, cpus_offset, cpu_model);
224             if (offset < 0) {
225                 return offset;
226             }
227         }
228 
229         ret = fdt_setprop(fdt, offset, "ibm,pft-size",
230                           pft_size_prop, sizeof(pft_size_prop));
231         if (ret < 0) {
232             return ret;
233         }
234 
235         ret = spapr_fixup_cpu_numa_dt(fdt, offset, cs);
236         if (ret < 0) {
237             return ret;
238         }
239 
240         ret = spapr_fixup_cpu_smt_dt(fdt, offset, cpu,
241                                      ppc_get_compat_smt_threads(cpu));
242         if (ret < 0) {
243             return ret;
244         }
245     }
246     return ret;
247 }
248 
249 
250 static size_t create_page_sizes_prop(CPUPPCState *env, uint32_t *prop,
251                                      size_t maxsize)
252 {
253     size_t maxcells = maxsize / sizeof(uint32_t);
254     int i, j, count;
255     uint32_t *p = prop;
256 
257     for (i = 0; i < PPC_PAGE_SIZES_MAX_SZ; i++) {
258         struct ppc_one_seg_page_size *sps = &env->sps.sps[i];
259 
260         if (!sps->page_shift) {
261             break;
262         }
263         for (count = 0; count < PPC_PAGE_SIZES_MAX_SZ; count++) {
264             if (sps->enc[count].page_shift == 0) {
265                 break;
266             }
267         }
268         if ((p - prop) >= (maxcells - 3 - count * 2)) {
269             break;
270         }
271         *(p++) = cpu_to_be32(sps->page_shift);
272         *(p++) = cpu_to_be32(sps->slb_enc);
273         *(p++) = cpu_to_be32(count);
274         for (j = 0; j < count; j++) {
275             *(p++) = cpu_to_be32(sps->enc[j].page_shift);
276             *(p++) = cpu_to_be32(sps->enc[j].pte_enc);
277         }
278     }
279 
280     return (p - prop) * sizeof(uint32_t);
281 }
282 
283 static hwaddr spapr_node0_size(void)
284 {
285     MachineState *machine = MACHINE(qdev_get_machine());
286 
287     if (nb_numa_nodes) {
288         int i;
289         for (i = 0; i < nb_numa_nodes; ++i) {
290             if (numa_info[i].node_mem) {
291                 return MIN(pow2floor(numa_info[i].node_mem),
292                            machine->ram_size);
293             }
294         }
295     }
296     return machine->ram_size;
297 }
298 
299 #define _FDT(exp) \
300     do { \
301         int ret = (exp);                                           \
302         if (ret < 0) {                                             \
303             fprintf(stderr, "qemu: error creating device tree: %s: %s\n", \
304                     #exp, fdt_strerror(ret));                      \
305             exit(1);                                               \
306         }                                                          \
307     } while (0)
308 
309 static void add_str(GString *s, const gchar *s1)
310 {
311     g_string_append_len(s, s1, strlen(s1) + 1);
312 }
313 
314 static void *spapr_create_fdt_skel(hwaddr initrd_base,
315                                    hwaddr initrd_size,
316                                    hwaddr kernel_size,
317                                    bool little_endian,
318                                    const char *kernel_cmdline,
319                                    uint32_t epow_irq)
320 {
321     void *fdt;
322     uint32_t start_prop = cpu_to_be32(initrd_base);
323     uint32_t end_prop = cpu_to_be32(initrd_base + initrd_size);
324     GString *hypertas = g_string_sized_new(256);
325     GString *qemu_hypertas = g_string_sized_new(256);
326     uint32_t refpoints[] = {cpu_to_be32(0x4), cpu_to_be32(0x4)};
327     uint32_t interrupt_server_ranges_prop[] = {0, cpu_to_be32(max_cpus)};
328     unsigned char vec5[] = {0x0, 0x0, 0x0, 0x0, 0x0, 0x80};
329     char *buf;
330 
331     add_str(hypertas, "hcall-pft");
332     add_str(hypertas, "hcall-term");
333     add_str(hypertas, "hcall-dabr");
334     add_str(hypertas, "hcall-interrupt");
335     add_str(hypertas, "hcall-tce");
336     add_str(hypertas, "hcall-vio");
337     add_str(hypertas, "hcall-splpar");
338     add_str(hypertas, "hcall-bulk");
339     add_str(hypertas, "hcall-set-mode");
340     add_str(qemu_hypertas, "hcall-memop1");
341 
342     fdt = g_malloc0(FDT_MAX_SIZE);
343     _FDT((fdt_create(fdt, FDT_MAX_SIZE)));
344 
345     if (kernel_size) {
346         _FDT((fdt_add_reservemap_entry(fdt, KERNEL_LOAD_ADDR, kernel_size)));
347     }
348     if (initrd_size) {
349         _FDT((fdt_add_reservemap_entry(fdt, initrd_base, initrd_size)));
350     }
351     _FDT((fdt_finish_reservemap(fdt)));
352 
353     /* Root node */
354     _FDT((fdt_begin_node(fdt, "")));
355     _FDT((fdt_property_string(fdt, "device_type", "chrp")));
356     _FDT((fdt_property_string(fdt, "model", "IBM pSeries (emulated by qemu)")));
357     _FDT((fdt_property_string(fdt, "compatible", "qemu,pseries")));
358 
359     /*
360      * Add info to guest to indentify which host is it being run on
361      * and what is the uuid of the guest
362      */
363     if (kvmppc_get_host_model(&buf)) {
364         _FDT((fdt_property_string(fdt, "host-model", buf)));
365         g_free(buf);
366     }
367     if (kvmppc_get_host_serial(&buf)) {
368         _FDT((fdt_property_string(fdt, "host-serial", buf)));
369         g_free(buf);
370     }
371 
372     buf = g_strdup_printf(UUID_FMT, qemu_uuid[0], qemu_uuid[1],
373                           qemu_uuid[2], qemu_uuid[3], qemu_uuid[4],
374                           qemu_uuid[5], qemu_uuid[6], qemu_uuid[7],
375                           qemu_uuid[8], qemu_uuid[9], qemu_uuid[10],
376                           qemu_uuid[11], qemu_uuid[12], qemu_uuid[13],
377                           qemu_uuid[14], qemu_uuid[15]);
378 
379     _FDT((fdt_property_string(fdt, "vm,uuid", buf)));
380     if (qemu_uuid_set) {
381         _FDT((fdt_property_string(fdt, "system-id", buf)));
382     }
383     g_free(buf);
384 
385     if (qemu_get_vm_name()) {
386         _FDT((fdt_property_string(fdt, "ibm,partition-name",
387                                   qemu_get_vm_name())));
388     }
389 
390     _FDT((fdt_property_cell(fdt, "#address-cells", 0x2)));
391     _FDT((fdt_property_cell(fdt, "#size-cells", 0x2)));
392 
393     /* /chosen */
394     _FDT((fdt_begin_node(fdt, "chosen")));
395 
396     /* Set Form1_affinity */
397     _FDT((fdt_property(fdt, "ibm,architecture-vec-5", vec5, sizeof(vec5))));
398 
399     _FDT((fdt_property_string(fdt, "bootargs", kernel_cmdline)));
400     _FDT((fdt_property(fdt, "linux,initrd-start",
401                        &start_prop, sizeof(start_prop))));
402     _FDT((fdt_property(fdt, "linux,initrd-end",
403                        &end_prop, sizeof(end_prop))));
404     if (kernel_size) {
405         uint64_t kprop[2] = { cpu_to_be64(KERNEL_LOAD_ADDR),
406                               cpu_to_be64(kernel_size) };
407 
408         _FDT((fdt_property(fdt, "qemu,boot-kernel", &kprop, sizeof(kprop))));
409         if (little_endian) {
410             _FDT((fdt_property(fdt, "qemu,boot-kernel-le", NULL, 0)));
411         }
412     }
413     if (boot_menu) {
414         _FDT((fdt_property_cell(fdt, "qemu,boot-menu", boot_menu)));
415     }
416     _FDT((fdt_property_cell(fdt, "qemu,graphic-width", graphic_width)));
417     _FDT((fdt_property_cell(fdt, "qemu,graphic-height", graphic_height)));
418     _FDT((fdt_property_cell(fdt, "qemu,graphic-depth", graphic_depth)));
419 
420     _FDT((fdt_end_node(fdt)));
421 
422     /* RTAS */
423     _FDT((fdt_begin_node(fdt, "rtas")));
424 
425     if (!kvm_enabled() || kvmppc_spapr_use_multitce()) {
426         add_str(hypertas, "hcall-multi-tce");
427     }
428     _FDT((fdt_property(fdt, "ibm,hypertas-functions", hypertas->str,
429                        hypertas->len)));
430     g_string_free(hypertas, TRUE);
431     _FDT((fdt_property(fdt, "qemu,hypertas-functions", qemu_hypertas->str,
432                        qemu_hypertas->len)));
433     g_string_free(qemu_hypertas, TRUE);
434 
435     _FDT((fdt_property(fdt, "ibm,associativity-reference-points",
436         refpoints, sizeof(refpoints))));
437 
438     _FDT((fdt_property_cell(fdt, "rtas-error-log-max", RTAS_ERROR_LOG_MAX)));
439     _FDT((fdt_property_cell(fdt, "rtas-event-scan-rate",
440                             RTAS_EVENT_SCAN_RATE)));
441 
442     if (msi_nonbroken) {
443         _FDT((fdt_property(fdt, "ibm,change-msix-capable", NULL, 0)));
444     }
445 
446     /*
447      * According to PAPR, rtas ibm,os-term does not guarantee a return
448      * back to the guest cpu.
449      *
450      * While an additional ibm,extended-os-term property indicates that
451      * rtas call return will always occur. Set this property.
452      */
453     _FDT((fdt_property(fdt, "ibm,extended-os-term", NULL, 0)));
454 
455     _FDT((fdt_end_node(fdt)));
456 
457     /* interrupt controller */
458     _FDT((fdt_begin_node(fdt, "interrupt-controller")));
459 
460     _FDT((fdt_property_string(fdt, "device_type",
461                               "PowerPC-External-Interrupt-Presentation")));
462     _FDT((fdt_property_string(fdt, "compatible", "IBM,ppc-xicp")));
463     _FDT((fdt_property(fdt, "interrupt-controller", NULL, 0)));
464     _FDT((fdt_property(fdt, "ibm,interrupt-server-ranges",
465                        interrupt_server_ranges_prop,
466                        sizeof(interrupt_server_ranges_prop))));
467     _FDT((fdt_property_cell(fdt, "#interrupt-cells", 2)));
468     _FDT((fdt_property_cell(fdt, "linux,phandle", PHANDLE_XICP)));
469     _FDT((fdt_property_cell(fdt, "phandle", PHANDLE_XICP)));
470 
471     _FDT((fdt_end_node(fdt)));
472 
473     /* vdevice */
474     _FDT((fdt_begin_node(fdt, "vdevice")));
475 
476     _FDT((fdt_property_string(fdt, "device_type", "vdevice")));
477     _FDT((fdt_property_string(fdt, "compatible", "IBM,vdevice")));
478     _FDT((fdt_property_cell(fdt, "#address-cells", 0x1)));
479     _FDT((fdt_property_cell(fdt, "#size-cells", 0x0)));
480     _FDT((fdt_property_cell(fdt, "#interrupt-cells", 0x2)));
481     _FDT((fdt_property(fdt, "interrupt-controller", NULL, 0)));
482 
483     _FDT((fdt_end_node(fdt)));
484 
485     /* event-sources */
486     spapr_events_fdt_skel(fdt, epow_irq);
487 
488     /* /hypervisor node */
489     if (kvm_enabled()) {
490         uint8_t hypercall[16];
491 
492         /* indicate KVM hypercall interface */
493         _FDT((fdt_begin_node(fdt, "hypervisor")));
494         _FDT((fdt_property_string(fdt, "compatible", "linux,kvm")));
495         if (kvmppc_has_cap_fixup_hcalls()) {
496             /*
497              * Older KVM versions with older guest kernels were broken with the
498              * magic page, don't allow the guest to map it.
499              */
500             kvmppc_get_hypercall(first_cpu->env_ptr, hypercall,
501                                  sizeof(hypercall));
502             _FDT((fdt_property(fdt, "hcall-instructions", hypercall,
503                               sizeof(hypercall))));
504         }
505         _FDT((fdt_end_node(fdt)));
506     }
507 
508     _FDT((fdt_end_node(fdt))); /* close root node */
509     _FDT((fdt_finish(fdt)));
510 
511     return fdt;
512 }
513 
514 static int spapr_populate_memory_node(void *fdt, int nodeid, hwaddr start,
515                                        hwaddr size)
516 {
517     uint32_t associativity[] = {
518         cpu_to_be32(0x4), /* length */
519         cpu_to_be32(0x0), cpu_to_be32(0x0),
520         cpu_to_be32(0x0), cpu_to_be32(nodeid)
521     };
522     char mem_name[32];
523     uint64_t mem_reg_property[2];
524     int off;
525 
526     mem_reg_property[0] = cpu_to_be64(start);
527     mem_reg_property[1] = cpu_to_be64(size);
528 
529     sprintf(mem_name, "memory@" TARGET_FMT_lx, start);
530     off = fdt_add_subnode(fdt, 0, mem_name);
531     _FDT(off);
532     _FDT((fdt_setprop_string(fdt, off, "device_type", "memory")));
533     _FDT((fdt_setprop(fdt, off, "reg", mem_reg_property,
534                       sizeof(mem_reg_property))));
535     _FDT((fdt_setprop(fdt, off, "ibm,associativity", associativity,
536                       sizeof(associativity))));
537     return off;
538 }
539 
540 static int spapr_populate_memory(sPAPRMachineState *spapr, void *fdt)
541 {
542     MachineState *machine = MACHINE(spapr);
543     hwaddr mem_start, node_size;
544     int i, nb_nodes = nb_numa_nodes;
545     NodeInfo *nodes = numa_info;
546     NodeInfo ramnode;
547 
548     /* No NUMA nodes, assume there is just one node with whole RAM */
549     if (!nb_numa_nodes) {
550         nb_nodes = 1;
551         ramnode.node_mem = machine->ram_size;
552         nodes = &ramnode;
553     }
554 
555     for (i = 0, mem_start = 0; i < nb_nodes; ++i) {
556         if (!nodes[i].node_mem) {
557             continue;
558         }
559         if (mem_start >= machine->ram_size) {
560             node_size = 0;
561         } else {
562             node_size = nodes[i].node_mem;
563             if (node_size > machine->ram_size - mem_start) {
564                 node_size = machine->ram_size - mem_start;
565             }
566         }
567         if (!mem_start) {
568             /* ppc_spapr_init() checks for rma_size <= node0_size already */
569             spapr_populate_memory_node(fdt, i, 0, spapr->rma_size);
570             mem_start += spapr->rma_size;
571             node_size -= spapr->rma_size;
572         }
573         for ( ; node_size; ) {
574             hwaddr sizetmp = pow2floor(node_size);
575 
576             /* mem_start != 0 here */
577             if (ctzl(mem_start) < ctzl(sizetmp)) {
578                 sizetmp = 1ULL << ctzl(mem_start);
579             }
580 
581             spapr_populate_memory_node(fdt, i, mem_start, sizetmp);
582             node_size -= sizetmp;
583             mem_start += sizetmp;
584         }
585     }
586 
587     return 0;
588 }
589 
590 static void spapr_populate_cpu_dt(CPUState *cs, void *fdt, int offset,
591                                   sPAPRMachineState *spapr)
592 {
593     PowerPCCPU *cpu = POWERPC_CPU(cs);
594     CPUPPCState *env = &cpu->env;
595     PowerPCCPUClass *pcc = POWERPC_CPU_GET_CLASS(cs);
596     int index = ppc_get_vcpu_dt_id(cpu);
597     uint32_t segs[] = {cpu_to_be32(28), cpu_to_be32(40),
598                        0xffffffff, 0xffffffff};
599     uint32_t tbfreq = kvm_enabled() ? kvmppc_get_tbfreq() : TIMEBASE_FREQ;
600     uint32_t cpufreq = kvm_enabled() ? kvmppc_get_clockfreq() : 1000000000;
601     uint32_t page_sizes_prop[64];
602     size_t page_sizes_prop_size;
603     uint32_t vcpus_per_socket = smp_threads * smp_cores;
604     uint32_t pft_size_prop[] = {0, cpu_to_be32(spapr->htab_shift)};
605 
606     /* Note: we keep CI large pages off for now because a 64K capable guest
607      * provisioned with large pages might otherwise try to map a qemu
608      * framebuffer (or other kind of memory mapped PCI BAR) using 64K pages
609      * even if that qemu runs on a 4k host.
610      *
611      * We can later add this bit back when we are confident this is not
612      * an issue (!HV KVM or 64K host)
613      */
614     uint8_t pa_features_206[] = { 6, 0,
615         0xf6, 0x1f, 0xc7, 0x00, 0x80, 0xc0 };
616     uint8_t pa_features_207[] = { 24, 0,
617         0xf6, 0x1f, 0xc7, 0xc0, 0x80, 0xf0,
618         0x80, 0x00, 0x00, 0x00, 0x00, 0x00,
619         0x00, 0x00, 0x00, 0x00, 0x80, 0x00,
620         0x80, 0x00, 0x80, 0x00, 0x80, 0x00 };
621     uint8_t *pa_features;
622     size_t pa_size;
623 
624     _FDT((fdt_setprop_cell(fdt, offset, "reg", index)));
625     _FDT((fdt_setprop_string(fdt, offset, "device_type", "cpu")));
626 
627     _FDT((fdt_setprop_cell(fdt, offset, "cpu-version", env->spr[SPR_PVR])));
628     _FDT((fdt_setprop_cell(fdt, offset, "d-cache-block-size",
629                            env->dcache_line_size)));
630     _FDT((fdt_setprop_cell(fdt, offset, "d-cache-line-size",
631                            env->dcache_line_size)));
632     _FDT((fdt_setprop_cell(fdt, offset, "i-cache-block-size",
633                            env->icache_line_size)));
634     _FDT((fdt_setprop_cell(fdt, offset, "i-cache-line-size",
635                            env->icache_line_size)));
636 
637     if (pcc->l1_dcache_size) {
638         _FDT((fdt_setprop_cell(fdt, offset, "d-cache-size",
639                                pcc->l1_dcache_size)));
640     } else {
641         fprintf(stderr, "Warning: Unknown L1 dcache size for cpu\n");
642     }
643     if (pcc->l1_icache_size) {
644         _FDT((fdt_setprop_cell(fdt, offset, "i-cache-size",
645                                pcc->l1_icache_size)));
646     } else {
647         fprintf(stderr, "Warning: Unknown L1 icache size for cpu\n");
648     }
649 
650     _FDT((fdt_setprop_cell(fdt, offset, "timebase-frequency", tbfreq)));
651     _FDT((fdt_setprop_cell(fdt, offset, "clock-frequency", cpufreq)));
652     _FDT((fdt_setprop_cell(fdt, offset, "slb-size", env->slb_nr)));
653     _FDT((fdt_setprop_cell(fdt, offset, "ibm,slb-size", env->slb_nr)));
654     _FDT((fdt_setprop_string(fdt, offset, "status", "okay")));
655     _FDT((fdt_setprop(fdt, offset, "64-bit", NULL, 0)));
656 
657     if (env->spr_cb[SPR_PURR].oea_read) {
658         _FDT((fdt_setprop(fdt, offset, "ibm,purr", NULL, 0)));
659     }
660 
661     if (env->mmu_model & POWERPC_MMU_1TSEG) {
662         _FDT((fdt_setprop(fdt, offset, "ibm,processor-segment-sizes",
663                           segs, sizeof(segs))));
664     }
665 
666     /* Advertise VMX/VSX (vector extensions) if available
667      *   0 / no property == no vector extensions
668      *   1               == VMX / Altivec available
669      *   2               == VSX available */
670     if (env->insns_flags & PPC_ALTIVEC) {
671         uint32_t vmx = (env->insns_flags2 & PPC2_VSX) ? 2 : 1;
672 
673         _FDT((fdt_setprop_cell(fdt, offset, "ibm,vmx", vmx)));
674     }
675 
676     /* Advertise DFP (Decimal Floating Point) if available
677      *   0 / no property == no DFP
678      *   1               == DFP available */
679     if (env->insns_flags2 & PPC2_DFP) {
680         _FDT((fdt_setprop_cell(fdt, offset, "ibm,dfp", 1)));
681     }
682 
683     page_sizes_prop_size = create_page_sizes_prop(env, page_sizes_prop,
684                                                   sizeof(page_sizes_prop));
685     if (page_sizes_prop_size) {
686         _FDT((fdt_setprop(fdt, offset, "ibm,segment-page-sizes",
687                           page_sizes_prop, page_sizes_prop_size)));
688     }
689 
690     /* Do the ibm,pa-features property, adjust it for ci-large-pages */
691     if (env->mmu_model == POWERPC_MMU_2_06) {
692         pa_features = pa_features_206;
693         pa_size = sizeof(pa_features_206);
694     } else /* env->mmu_model == POWERPC_MMU_2_07 */ {
695         pa_features = pa_features_207;
696         pa_size = sizeof(pa_features_207);
697     }
698     if (env->ci_large_pages) {
699         pa_features[3] |= 0x20;
700     }
701     _FDT((fdt_setprop(fdt, offset, "ibm,pa-features", pa_features, pa_size)));
702 
703     _FDT((fdt_setprop_cell(fdt, offset, "ibm,chip-id",
704                            cs->cpu_index / vcpus_per_socket)));
705 
706     _FDT((fdt_setprop(fdt, offset, "ibm,pft-size",
707                       pft_size_prop, sizeof(pft_size_prop))));
708 
709     _FDT(spapr_fixup_cpu_numa_dt(fdt, offset, cs));
710 
711     _FDT(spapr_fixup_cpu_smt_dt(fdt, offset, cpu,
712                                 ppc_get_compat_smt_threads(cpu)));
713 }
714 
715 static void spapr_populate_cpus_dt_node(void *fdt, sPAPRMachineState *spapr)
716 {
717     CPUState *cs;
718     int cpus_offset;
719     char *nodename;
720     int smt = kvmppc_smt_threads();
721 
722     cpus_offset = fdt_add_subnode(fdt, 0, "cpus");
723     _FDT(cpus_offset);
724     _FDT((fdt_setprop_cell(fdt, cpus_offset, "#address-cells", 0x1)));
725     _FDT((fdt_setprop_cell(fdt, cpus_offset, "#size-cells", 0x0)));
726 
727     /*
728      * We walk the CPUs in reverse order to ensure that CPU DT nodes
729      * created by fdt_add_subnode() end up in the right order in FDT
730      * for the guest kernel the enumerate the CPUs correctly.
731      */
732     CPU_FOREACH_REVERSE(cs) {
733         PowerPCCPU *cpu = POWERPC_CPU(cs);
734         int index = ppc_get_vcpu_dt_id(cpu);
735         DeviceClass *dc = DEVICE_GET_CLASS(cs);
736         int offset;
737 
738         if ((index % smt) != 0) {
739             continue;
740         }
741 
742         nodename = g_strdup_printf("%s@%x", dc->fw_name, index);
743         offset = fdt_add_subnode(fdt, cpus_offset, nodename);
744         g_free(nodename);
745         _FDT(offset);
746         spapr_populate_cpu_dt(cs, fdt, offset, spapr);
747     }
748 
749 }
750 
751 /*
752  * Adds ibm,dynamic-reconfiguration-memory node.
753  * Refer to docs/specs/ppc-spapr-hotplug.txt for the documentation
754  * of this device tree node.
755  */
756 static int spapr_populate_drconf_memory(sPAPRMachineState *spapr, void *fdt)
757 {
758     MachineState *machine = MACHINE(spapr);
759     int ret, i, offset;
760     uint64_t lmb_size = SPAPR_MEMORY_BLOCK_SIZE;
761     uint32_t prop_lmb_size[] = {0, cpu_to_be32(lmb_size)};
762     uint32_t nr_lmbs = (machine->maxram_size - machine->ram_size)/lmb_size;
763     uint32_t *int_buf, *cur_index, buf_len;
764     int nr_nodes = nb_numa_nodes ? nb_numa_nodes : 1;
765 
766     /*
767      * Don't create the node if there are no DR LMBs.
768      */
769     if (!nr_lmbs) {
770         return 0;
771     }
772 
773     /*
774      * Allocate enough buffer size to fit in ibm,dynamic-memory
775      * or ibm,associativity-lookup-arrays
776      */
777     buf_len = MAX(nr_lmbs * SPAPR_DR_LMB_LIST_ENTRY_SIZE + 1, nr_nodes * 4 + 2)
778               * sizeof(uint32_t);
779     cur_index = int_buf = g_malloc0(buf_len);
780 
781     offset = fdt_add_subnode(fdt, 0, "ibm,dynamic-reconfiguration-memory");
782 
783     ret = fdt_setprop(fdt, offset, "ibm,lmb-size", prop_lmb_size,
784                     sizeof(prop_lmb_size));
785     if (ret < 0) {
786         goto out;
787     }
788 
789     ret = fdt_setprop_cell(fdt, offset, "ibm,memory-flags-mask", 0xff);
790     if (ret < 0) {
791         goto out;
792     }
793 
794     ret = fdt_setprop_cell(fdt, offset, "ibm,memory-preservation-time", 0x0);
795     if (ret < 0) {
796         goto out;
797     }
798 
799     /* ibm,dynamic-memory */
800     int_buf[0] = cpu_to_be32(nr_lmbs);
801     cur_index++;
802     for (i = 0; i < nr_lmbs; i++) {
803         sPAPRDRConnector *drc;
804         sPAPRDRConnectorClass *drck;
805         uint64_t addr = i * lmb_size + spapr->hotplug_memory.base;;
806         uint32_t *dynamic_memory = cur_index;
807 
808         drc = spapr_dr_connector_by_id(SPAPR_DR_CONNECTOR_TYPE_LMB,
809                                        addr/lmb_size);
810         g_assert(drc);
811         drck = SPAPR_DR_CONNECTOR_GET_CLASS(drc);
812 
813         dynamic_memory[0] = cpu_to_be32(addr >> 32);
814         dynamic_memory[1] = cpu_to_be32(addr & 0xffffffff);
815         dynamic_memory[2] = cpu_to_be32(drck->get_index(drc));
816         dynamic_memory[3] = cpu_to_be32(0); /* reserved */
817         dynamic_memory[4] = cpu_to_be32(numa_get_node(addr, NULL));
818         if (addr < machine->ram_size ||
819                     memory_region_present(get_system_memory(), addr)) {
820             dynamic_memory[5] = cpu_to_be32(SPAPR_LMB_FLAGS_ASSIGNED);
821         } else {
822             dynamic_memory[5] = cpu_to_be32(0);
823         }
824 
825         cur_index += SPAPR_DR_LMB_LIST_ENTRY_SIZE;
826     }
827     ret = fdt_setprop(fdt, offset, "ibm,dynamic-memory", int_buf, buf_len);
828     if (ret < 0) {
829         goto out;
830     }
831 
832     /* ibm,associativity-lookup-arrays */
833     cur_index = int_buf;
834     int_buf[0] = cpu_to_be32(nr_nodes);
835     int_buf[1] = cpu_to_be32(4); /* Number of entries per associativity list */
836     cur_index += 2;
837     for (i = 0; i < nr_nodes; i++) {
838         uint32_t associativity[] = {
839             cpu_to_be32(0x0),
840             cpu_to_be32(0x0),
841             cpu_to_be32(0x0),
842             cpu_to_be32(i)
843         };
844         memcpy(cur_index, associativity, sizeof(associativity));
845         cur_index += 4;
846     }
847     ret = fdt_setprop(fdt, offset, "ibm,associativity-lookup-arrays", int_buf,
848             (cur_index - int_buf) * sizeof(uint32_t));
849 out:
850     g_free(int_buf);
851     return ret;
852 }
853 
854 int spapr_h_cas_compose_response(sPAPRMachineState *spapr,
855                                  target_ulong addr, target_ulong size,
856                                  bool cpu_update, bool memory_update)
857 {
858     void *fdt, *fdt_skel;
859     sPAPRDeviceTreeUpdateHeader hdr = { .version_id = 1 };
860     sPAPRMachineClass *smc = SPAPR_MACHINE_GET_CLASS(qdev_get_machine());
861 
862     size -= sizeof(hdr);
863 
864     /* Create sceleton */
865     fdt_skel = g_malloc0(size);
866     _FDT((fdt_create(fdt_skel, size)));
867     _FDT((fdt_begin_node(fdt_skel, "")));
868     _FDT((fdt_end_node(fdt_skel)));
869     _FDT((fdt_finish(fdt_skel)));
870     fdt = g_malloc0(size);
871     _FDT((fdt_open_into(fdt_skel, fdt, size)));
872     g_free(fdt_skel);
873 
874     /* Fixup cpu nodes */
875     if (cpu_update) {
876         _FDT((spapr_fixup_cpu_dt(fdt, spapr)));
877     }
878 
879     /* Generate ibm,dynamic-reconfiguration-memory node if required */
880     if (memory_update && smc->dr_lmb_enabled) {
881         _FDT((spapr_populate_drconf_memory(spapr, fdt)));
882     }
883 
884     /* Pack resulting tree */
885     _FDT((fdt_pack(fdt)));
886 
887     if (fdt_totalsize(fdt) + sizeof(hdr) > size) {
888         trace_spapr_cas_failed(size);
889         return -1;
890     }
891 
892     cpu_physical_memory_write(addr, &hdr, sizeof(hdr));
893     cpu_physical_memory_write(addr + sizeof(hdr), fdt, fdt_totalsize(fdt));
894     trace_spapr_cas_continue(fdt_totalsize(fdt) + sizeof(hdr));
895     g_free(fdt);
896 
897     return 0;
898 }
899 
900 static void spapr_finalize_fdt(sPAPRMachineState *spapr,
901                                hwaddr fdt_addr,
902                                hwaddr rtas_addr,
903                                hwaddr rtas_size)
904 {
905     MachineState *machine = MACHINE(qdev_get_machine());
906     sPAPRMachineClass *smc = SPAPR_MACHINE_GET_CLASS(machine);
907     const char *boot_device = machine->boot_order;
908     int ret, i;
909     size_t cb = 0;
910     char *bootlist;
911     void *fdt;
912     sPAPRPHBState *phb;
913 
914     fdt = g_malloc(FDT_MAX_SIZE);
915 
916     /* open out the base tree into a temp buffer for the final tweaks */
917     _FDT((fdt_open_into(spapr->fdt_skel, fdt, FDT_MAX_SIZE)));
918 
919     ret = spapr_populate_memory(spapr, fdt);
920     if (ret < 0) {
921         fprintf(stderr, "couldn't setup memory nodes in fdt\n");
922         exit(1);
923     }
924 
925     ret = spapr_populate_vdevice(spapr->vio_bus, fdt);
926     if (ret < 0) {
927         fprintf(stderr, "couldn't setup vio devices in fdt\n");
928         exit(1);
929     }
930 
931     if (object_resolve_path_type("", TYPE_SPAPR_RNG, NULL)) {
932         ret = spapr_rng_populate_dt(fdt);
933         if (ret < 0) {
934             fprintf(stderr, "could not set up rng device in the fdt\n");
935             exit(1);
936         }
937     }
938 
939     QLIST_FOREACH(phb, &spapr->phbs, list) {
940         ret = spapr_populate_pci_dt(phb, PHANDLE_XICP, fdt);
941     }
942 
943     if (ret < 0) {
944         fprintf(stderr, "couldn't setup PCI devices in fdt\n");
945         exit(1);
946     }
947 
948     /* RTAS */
949     ret = spapr_rtas_device_tree_setup(fdt, rtas_addr, rtas_size);
950     if (ret < 0) {
951         fprintf(stderr, "Couldn't set up RTAS device tree properties\n");
952     }
953 
954     /* cpus */
955     spapr_populate_cpus_dt_node(fdt, spapr);
956 
957     bootlist = get_boot_devices_list(&cb, true);
958     if (cb && bootlist) {
959         int offset = fdt_path_offset(fdt, "/chosen");
960         if (offset < 0) {
961             exit(1);
962         }
963         for (i = 0; i < cb; i++) {
964             if (bootlist[i] == '\n') {
965                 bootlist[i] = ' ';
966             }
967 
968         }
969         ret = fdt_setprop_string(fdt, offset, "qemu,boot-list", bootlist);
970     }
971 
972     if (boot_device && strlen(boot_device)) {
973         int offset = fdt_path_offset(fdt, "/chosen");
974 
975         if (offset < 0) {
976             exit(1);
977         }
978         fdt_setprop_string(fdt, offset, "qemu,boot-device", boot_device);
979     }
980 
981     if (!spapr->has_graphics) {
982         spapr_populate_chosen_stdout(fdt, spapr->vio_bus);
983     }
984 
985     if (smc->dr_lmb_enabled) {
986         _FDT(spapr_drc_populate_dt(fdt, 0, NULL, SPAPR_DR_CONNECTOR_TYPE_LMB));
987     }
988 
989     _FDT((fdt_pack(fdt)));
990 
991     if (fdt_totalsize(fdt) > FDT_MAX_SIZE) {
992         error_report("FDT too big ! 0x%x bytes (max is 0x%x)",
993                      fdt_totalsize(fdt), FDT_MAX_SIZE);
994         exit(1);
995     }
996 
997     qemu_fdt_dumpdtb(fdt, fdt_totalsize(fdt));
998     cpu_physical_memory_write(fdt_addr, fdt, fdt_totalsize(fdt));
999 
1000     g_free(bootlist);
1001     g_free(fdt);
1002 }
1003 
1004 static uint64_t translate_kernel_address(void *opaque, uint64_t addr)
1005 {
1006     return (addr & 0x0fffffff) + KERNEL_LOAD_ADDR;
1007 }
1008 
1009 static void emulate_spapr_hypercall(PowerPCCPU *cpu)
1010 {
1011     CPUPPCState *env = &cpu->env;
1012 
1013     if (msr_pr) {
1014         hcall_dprintf("Hypercall made with MSR[PR]=1\n");
1015         env->gpr[3] = H_PRIVILEGE;
1016     } else {
1017         env->gpr[3] = spapr_hypercall(cpu, env->gpr[3], &env->gpr[4]);
1018     }
1019 }
1020 
1021 #define HPTE(_table, _i)   (void *)(((uint64_t *)(_table)) + ((_i) * 2))
1022 #define HPTE_VALID(_hpte)  (tswap64(*((uint64_t *)(_hpte))) & HPTE64_V_VALID)
1023 #define HPTE_DIRTY(_hpte)  (tswap64(*((uint64_t *)(_hpte))) & HPTE64_V_HPTE_DIRTY)
1024 #define CLEAN_HPTE(_hpte)  ((*(uint64_t *)(_hpte)) &= tswap64(~HPTE64_V_HPTE_DIRTY))
1025 #define DIRTY_HPTE(_hpte)  ((*(uint64_t *)(_hpte)) |= tswap64(HPTE64_V_HPTE_DIRTY))
1026 
1027 /*
1028  * Get the fd to access the kernel htab, re-opening it if necessary
1029  */
1030 static int get_htab_fd(sPAPRMachineState *spapr)
1031 {
1032     if (spapr->htab_fd >= 0) {
1033         return spapr->htab_fd;
1034     }
1035 
1036     spapr->htab_fd = kvmppc_get_htab_fd(false);
1037     if (spapr->htab_fd < 0) {
1038         error_report("Unable to open fd for reading hash table from KVM: %s",
1039                      strerror(errno));
1040     }
1041 
1042     return spapr->htab_fd;
1043 }
1044 
1045 static void close_htab_fd(sPAPRMachineState *spapr)
1046 {
1047     if (spapr->htab_fd >= 0) {
1048         close(spapr->htab_fd);
1049     }
1050     spapr->htab_fd = -1;
1051 }
1052 
1053 static int spapr_hpt_shift_for_ramsize(uint64_t ramsize)
1054 {
1055     int shift;
1056 
1057     /* We aim for a hash table of size 1/128 the size of RAM (rounded
1058      * up).  The PAPR recommendation is actually 1/64 of RAM size, but
1059      * that's much more than is needed for Linux guests */
1060     shift = ctz64(pow2ceil(ramsize)) - 7;
1061     shift = MAX(shift, 18); /* Minimum architected size */
1062     shift = MIN(shift, 46); /* Maximum architected size */
1063     return shift;
1064 }
1065 
1066 static void spapr_reallocate_hpt(sPAPRMachineState *spapr, int shift,
1067                                  Error **errp)
1068 {
1069     long rc;
1070 
1071     /* Clean up any HPT info from a previous boot */
1072     g_free(spapr->htab);
1073     spapr->htab = NULL;
1074     spapr->htab_shift = 0;
1075     close_htab_fd(spapr);
1076 
1077     rc = kvmppc_reset_htab(shift);
1078     if (rc < 0) {
1079         /* kernel-side HPT needed, but couldn't allocate one */
1080         error_setg_errno(errp, errno,
1081                          "Failed to allocate KVM HPT of order %d (try smaller maxmem?)",
1082                          shift);
1083         /* This is almost certainly fatal, but if the caller really
1084          * wants to carry on with shift == 0, it's welcome to try */
1085     } else if (rc > 0) {
1086         /* kernel-side HPT allocated */
1087         if (rc != shift) {
1088             error_setg(errp,
1089                        "Requested order %d HPT, but kernel allocated order %ld (try smaller maxmem?)",
1090                        shift, rc);
1091         }
1092 
1093         spapr->htab_shift = shift;
1094         spapr->htab = NULL;
1095     } else {
1096         /* kernel-side HPT not needed, allocate in userspace instead */
1097         size_t size = 1ULL << shift;
1098         int i;
1099 
1100         spapr->htab = qemu_memalign(size, size);
1101         if (!spapr->htab) {
1102             error_setg_errno(errp, errno,
1103                              "Could not allocate HPT of order %d", shift);
1104             return;
1105         }
1106 
1107         memset(spapr->htab, 0, size);
1108         spapr->htab_shift = shift;
1109 
1110         for (i = 0; i < size / HASH_PTE_SIZE_64; i++) {
1111             DIRTY_HPTE(HPTE(spapr->htab, i));
1112         }
1113     }
1114 }
1115 
1116 static int find_unknown_sysbus_device(SysBusDevice *sbdev, void *opaque)
1117 {
1118     bool matched = false;
1119 
1120     if (object_dynamic_cast(OBJECT(sbdev), TYPE_SPAPR_PCI_HOST_BRIDGE)) {
1121         matched = true;
1122     }
1123 
1124     if (!matched) {
1125         error_report("Device %s is not supported by this machine yet.",
1126                      qdev_fw_name(DEVICE(sbdev)));
1127         exit(1);
1128     }
1129 
1130     return 0;
1131 }
1132 
1133 static void ppc_spapr_reset(void)
1134 {
1135     MachineState *machine = MACHINE(qdev_get_machine());
1136     sPAPRMachineState *spapr = SPAPR_MACHINE(machine);
1137     PowerPCCPU *first_ppc_cpu;
1138     uint32_t rtas_limit;
1139 
1140     /* Check for unknown sysbus devices */
1141     foreach_dynamic_sysbus_device(find_unknown_sysbus_device, NULL);
1142 
1143     /* Allocate and/or reset the hash page table */
1144     spapr_reallocate_hpt(spapr,
1145                          spapr_hpt_shift_for_ramsize(machine->maxram_size),
1146                          &error_fatal);
1147 
1148     /* Update the RMA size if necessary */
1149     if (spapr->vrma_adjust) {
1150         spapr->rma_size = kvmppc_rma_size(spapr_node0_size(),
1151                                           spapr->htab_shift);
1152     }
1153 
1154     qemu_devices_reset();
1155 
1156     /*
1157      * We place the device tree and RTAS just below either the top of the RMA,
1158      * or just below 2GB, whichever is lowere, so that it can be
1159      * processed with 32-bit real mode code if necessary
1160      */
1161     rtas_limit = MIN(spapr->rma_size, RTAS_MAX_ADDR);
1162     spapr->rtas_addr = rtas_limit - RTAS_MAX_SIZE;
1163     spapr->fdt_addr = spapr->rtas_addr - FDT_MAX_SIZE;
1164 
1165     /* Load the fdt */
1166     spapr_finalize_fdt(spapr, spapr->fdt_addr, spapr->rtas_addr,
1167                        spapr->rtas_size);
1168 
1169     /* Copy RTAS over */
1170     cpu_physical_memory_write(spapr->rtas_addr, spapr->rtas_blob,
1171                               spapr->rtas_size);
1172 
1173     /* Set up the entry state */
1174     first_ppc_cpu = POWERPC_CPU(first_cpu);
1175     first_ppc_cpu->env.gpr[3] = spapr->fdt_addr;
1176     first_ppc_cpu->env.gpr[5] = 0;
1177     first_cpu->halted = 0;
1178     first_ppc_cpu->env.nip = SPAPR_ENTRY_POINT;
1179 
1180 }
1181 
1182 static void spapr_cpu_reset(void *opaque)
1183 {
1184     sPAPRMachineState *spapr = SPAPR_MACHINE(qdev_get_machine());
1185     PowerPCCPU *cpu = opaque;
1186     CPUState *cs = CPU(cpu);
1187     CPUPPCState *env = &cpu->env;
1188 
1189     cpu_reset(cs);
1190 
1191     /* All CPUs start halted.  CPU0 is unhalted from the machine level
1192      * reset code and the rest are explicitly started up by the guest
1193      * using an RTAS call */
1194     cs->halted = 1;
1195 
1196     env->spr[SPR_HIOR] = 0;
1197 
1198     ppc_hash64_set_external_hpt(cpu, spapr->htab, spapr->htab_shift,
1199                                 &error_fatal);
1200 }
1201 
1202 static void spapr_create_nvram(sPAPRMachineState *spapr)
1203 {
1204     DeviceState *dev = qdev_create(&spapr->vio_bus->bus, "spapr-nvram");
1205     DriveInfo *dinfo = drive_get(IF_PFLASH, 0, 0);
1206 
1207     if (dinfo) {
1208         qdev_prop_set_drive(dev, "drive", blk_by_legacy_dinfo(dinfo),
1209                             &error_fatal);
1210     }
1211 
1212     qdev_init_nofail(dev);
1213 
1214     spapr->nvram = (struct sPAPRNVRAM *)dev;
1215 }
1216 
1217 static void spapr_rtc_create(sPAPRMachineState *spapr)
1218 {
1219     DeviceState *dev = qdev_create(NULL, TYPE_SPAPR_RTC);
1220 
1221     qdev_init_nofail(dev);
1222     spapr->rtc = dev;
1223 
1224     object_property_add_alias(qdev_get_machine(), "rtc-time",
1225                               OBJECT(spapr->rtc), "date", NULL);
1226 }
1227 
1228 /* Returns whether we want to use VGA or not */
1229 static bool spapr_vga_init(PCIBus *pci_bus, Error **errp)
1230 {
1231     switch (vga_interface_type) {
1232     case VGA_NONE:
1233         return false;
1234     case VGA_DEVICE:
1235         return true;
1236     case VGA_STD:
1237     case VGA_VIRTIO:
1238         return pci_vga_init(pci_bus) != NULL;
1239     default:
1240         error_setg(errp,
1241                    "Unsupported VGA mode, only -vga std or -vga virtio is supported");
1242         return false;
1243     }
1244 }
1245 
1246 static int spapr_post_load(void *opaque, int version_id)
1247 {
1248     sPAPRMachineState *spapr = (sPAPRMachineState *)opaque;
1249     int err = 0;
1250 
1251     /* In earlier versions, there was no separate qdev for the PAPR
1252      * RTC, so the RTC offset was stored directly in sPAPREnvironment.
1253      * So when migrating from those versions, poke the incoming offset
1254      * value into the RTC device */
1255     if (version_id < 3) {
1256         err = spapr_rtc_import_offset(spapr->rtc, spapr->rtc_offset);
1257     }
1258 
1259     return err;
1260 }
1261 
1262 static bool version_before_3(void *opaque, int version_id)
1263 {
1264     return version_id < 3;
1265 }
1266 
1267 static const VMStateDescription vmstate_spapr = {
1268     .name = "spapr",
1269     .version_id = 3,
1270     .minimum_version_id = 1,
1271     .post_load = spapr_post_load,
1272     .fields = (VMStateField[]) {
1273         /* used to be @next_irq */
1274         VMSTATE_UNUSED_BUFFER(version_before_3, 0, 4),
1275 
1276         /* RTC offset */
1277         VMSTATE_UINT64_TEST(rtc_offset, sPAPRMachineState, version_before_3),
1278 
1279         VMSTATE_PPC_TIMEBASE_V(tb, sPAPRMachineState, 2),
1280         VMSTATE_END_OF_LIST()
1281     },
1282 };
1283 
1284 static int htab_save_setup(QEMUFile *f, void *opaque)
1285 {
1286     sPAPRMachineState *spapr = opaque;
1287 
1288     /* "Iteration" header */
1289     qemu_put_be32(f, spapr->htab_shift);
1290 
1291     if (spapr->htab) {
1292         spapr->htab_save_index = 0;
1293         spapr->htab_first_pass = true;
1294     } else {
1295         assert(kvm_enabled());
1296     }
1297 
1298 
1299     return 0;
1300 }
1301 
1302 static void htab_save_first_pass(QEMUFile *f, sPAPRMachineState *spapr,
1303                                  int64_t max_ns)
1304 {
1305     bool has_timeout = max_ns != -1;
1306     int htabslots = HTAB_SIZE(spapr) / HASH_PTE_SIZE_64;
1307     int index = spapr->htab_save_index;
1308     int64_t starttime = qemu_clock_get_ns(QEMU_CLOCK_REALTIME);
1309 
1310     assert(spapr->htab_first_pass);
1311 
1312     do {
1313         int chunkstart;
1314 
1315         /* Consume invalid HPTEs */
1316         while ((index < htabslots)
1317                && !HPTE_VALID(HPTE(spapr->htab, index))) {
1318             index++;
1319             CLEAN_HPTE(HPTE(spapr->htab, index));
1320         }
1321 
1322         /* Consume valid HPTEs */
1323         chunkstart = index;
1324         while ((index < htabslots) && (index - chunkstart < USHRT_MAX)
1325                && HPTE_VALID(HPTE(spapr->htab, index))) {
1326             index++;
1327             CLEAN_HPTE(HPTE(spapr->htab, index));
1328         }
1329 
1330         if (index > chunkstart) {
1331             int n_valid = index - chunkstart;
1332 
1333             qemu_put_be32(f, chunkstart);
1334             qemu_put_be16(f, n_valid);
1335             qemu_put_be16(f, 0);
1336             qemu_put_buffer(f, HPTE(spapr->htab, chunkstart),
1337                             HASH_PTE_SIZE_64 * n_valid);
1338 
1339             if (has_timeout &&
1340                 (qemu_clock_get_ns(QEMU_CLOCK_REALTIME) - starttime) > max_ns) {
1341                 break;
1342             }
1343         }
1344     } while ((index < htabslots) && !qemu_file_rate_limit(f));
1345 
1346     if (index >= htabslots) {
1347         assert(index == htabslots);
1348         index = 0;
1349         spapr->htab_first_pass = false;
1350     }
1351     spapr->htab_save_index = index;
1352 }
1353 
1354 static int htab_save_later_pass(QEMUFile *f, sPAPRMachineState *spapr,
1355                                 int64_t max_ns)
1356 {
1357     bool final = max_ns < 0;
1358     int htabslots = HTAB_SIZE(spapr) / HASH_PTE_SIZE_64;
1359     int examined = 0, sent = 0;
1360     int index = spapr->htab_save_index;
1361     int64_t starttime = qemu_clock_get_ns(QEMU_CLOCK_REALTIME);
1362 
1363     assert(!spapr->htab_first_pass);
1364 
1365     do {
1366         int chunkstart, invalidstart;
1367 
1368         /* Consume non-dirty HPTEs */
1369         while ((index < htabslots)
1370                && !HPTE_DIRTY(HPTE(spapr->htab, index))) {
1371             index++;
1372             examined++;
1373         }
1374 
1375         chunkstart = index;
1376         /* Consume valid dirty HPTEs */
1377         while ((index < htabslots) && (index - chunkstart < USHRT_MAX)
1378                && HPTE_DIRTY(HPTE(spapr->htab, index))
1379                && HPTE_VALID(HPTE(spapr->htab, index))) {
1380             CLEAN_HPTE(HPTE(spapr->htab, index));
1381             index++;
1382             examined++;
1383         }
1384 
1385         invalidstart = index;
1386         /* Consume invalid dirty HPTEs */
1387         while ((index < htabslots) && (index - invalidstart < USHRT_MAX)
1388                && HPTE_DIRTY(HPTE(spapr->htab, index))
1389                && !HPTE_VALID(HPTE(spapr->htab, index))) {
1390             CLEAN_HPTE(HPTE(spapr->htab, index));
1391             index++;
1392             examined++;
1393         }
1394 
1395         if (index > chunkstart) {
1396             int n_valid = invalidstart - chunkstart;
1397             int n_invalid = index - invalidstart;
1398 
1399             qemu_put_be32(f, chunkstart);
1400             qemu_put_be16(f, n_valid);
1401             qemu_put_be16(f, n_invalid);
1402             qemu_put_buffer(f, HPTE(spapr->htab, chunkstart),
1403                             HASH_PTE_SIZE_64 * n_valid);
1404             sent += index - chunkstart;
1405 
1406             if (!final && (qemu_clock_get_ns(QEMU_CLOCK_REALTIME) - starttime) > max_ns) {
1407                 break;
1408             }
1409         }
1410 
1411         if (examined >= htabslots) {
1412             break;
1413         }
1414 
1415         if (index >= htabslots) {
1416             assert(index == htabslots);
1417             index = 0;
1418         }
1419     } while ((examined < htabslots) && (!qemu_file_rate_limit(f) || final));
1420 
1421     if (index >= htabslots) {
1422         assert(index == htabslots);
1423         index = 0;
1424     }
1425 
1426     spapr->htab_save_index = index;
1427 
1428     return (examined >= htabslots) && (sent == 0) ? 1 : 0;
1429 }
1430 
1431 #define MAX_ITERATION_NS    5000000 /* 5 ms */
1432 #define MAX_KVM_BUF_SIZE    2048
1433 
1434 static int htab_save_iterate(QEMUFile *f, void *opaque)
1435 {
1436     sPAPRMachineState *spapr = opaque;
1437     int fd;
1438     int rc = 0;
1439 
1440     /* Iteration header */
1441     qemu_put_be32(f, 0);
1442 
1443     if (!spapr->htab) {
1444         assert(kvm_enabled());
1445 
1446         fd = get_htab_fd(spapr);
1447         if (fd < 0) {
1448             return fd;
1449         }
1450 
1451         rc = kvmppc_save_htab(f, fd, MAX_KVM_BUF_SIZE, MAX_ITERATION_NS);
1452         if (rc < 0) {
1453             return rc;
1454         }
1455     } else  if (spapr->htab_first_pass) {
1456         htab_save_first_pass(f, spapr, MAX_ITERATION_NS);
1457     } else {
1458         rc = htab_save_later_pass(f, spapr, MAX_ITERATION_NS);
1459     }
1460 
1461     /* End marker */
1462     qemu_put_be32(f, 0);
1463     qemu_put_be16(f, 0);
1464     qemu_put_be16(f, 0);
1465 
1466     return rc;
1467 }
1468 
1469 static int htab_save_complete(QEMUFile *f, void *opaque)
1470 {
1471     sPAPRMachineState *spapr = opaque;
1472     int fd;
1473 
1474     /* Iteration header */
1475     qemu_put_be32(f, 0);
1476 
1477     if (!spapr->htab) {
1478         int rc;
1479 
1480         assert(kvm_enabled());
1481 
1482         fd = get_htab_fd(spapr);
1483         if (fd < 0) {
1484             return fd;
1485         }
1486 
1487         rc = kvmppc_save_htab(f, fd, MAX_KVM_BUF_SIZE, -1);
1488         if (rc < 0) {
1489             return rc;
1490         }
1491         close_htab_fd(spapr);
1492     } else {
1493         if (spapr->htab_first_pass) {
1494             htab_save_first_pass(f, spapr, -1);
1495         }
1496         htab_save_later_pass(f, spapr, -1);
1497     }
1498 
1499     /* End marker */
1500     qemu_put_be32(f, 0);
1501     qemu_put_be16(f, 0);
1502     qemu_put_be16(f, 0);
1503 
1504     return 0;
1505 }
1506 
1507 static int htab_load(QEMUFile *f, void *opaque, int version_id)
1508 {
1509     sPAPRMachineState *spapr = opaque;
1510     uint32_t section_hdr;
1511     int fd = -1;
1512 
1513     if (version_id < 1 || version_id > 1) {
1514         error_report("htab_load() bad version");
1515         return -EINVAL;
1516     }
1517 
1518     section_hdr = qemu_get_be32(f);
1519 
1520     if (section_hdr) {
1521         Error *local_err = NULL;
1522 
1523         /* First section gives the htab size */
1524         spapr_reallocate_hpt(spapr, section_hdr, &local_err);
1525         if (local_err) {
1526             error_report_err(local_err);
1527             return -EINVAL;
1528         }
1529         return 0;
1530     }
1531 
1532     if (!spapr->htab) {
1533         assert(kvm_enabled());
1534 
1535         fd = kvmppc_get_htab_fd(true);
1536         if (fd < 0) {
1537             error_report("Unable to open fd to restore KVM hash table: %s",
1538                          strerror(errno));
1539         }
1540     }
1541 
1542     while (true) {
1543         uint32_t index;
1544         uint16_t n_valid, n_invalid;
1545 
1546         index = qemu_get_be32(f);
1547         n_valid = qemu_get_be16(f);
1548         n_invalid = qemu_get_be16(f);
1549 
1550         if ((index == 0) && (n_valid == 0) && (n_invalid == 0)) {
1551             /* End of Stream */
1552             break;
1553         }
1554 
1555         if ((index + n_valid + n_invalid) >
1556             (HTAB_SIZE(spapr) / HASH_PTE_SIZE_64)) {
1557             /* Bad index in stream */
1558             error_report(
1559                 "htab_load() bad index %d (%hd+%hd entries) in htab stream (htab_shift=%d)",
1560                 index, n_valid, n_invalid, spapr->htab_shift);
1561             return -EINVAL;
1562         }
1563 
1564         if (spapr->htab) {
1565             if (n_valid) {
1566                 qemu_get_buffer(f, HPTE(spapr->htab, index),
1567                                 HASH_PTE_SIZE_64 * n_valid);
1568             }
1569             if (n_invalid) {
1570                 memset(HPTE(spapr->htab, index + n_valid), 0,
1571                        HASH_PTE_SIZE_64 * n_invalid);
1572             }
1573         } else {
1574             int rc;
1575 
1576             assert(fd >= 0);
1577 
1578             rc = kvmppc_load_htab_chunk(f, fd, index, n_valid, n_invalid);
1579             if (rc < 0) {
1580                 return rc;
1581             }
1582         }
1583     }
1584 
1585     if (!spapr->htab) {
1586         assert(fd >= 0);
1587         close(fd);
1588     }
1589 
1590     return 0;
1591 }
1592 
1593 static SaveVMHandlers savevm_htab_handlers = {
1594     .save_live_setup = htab_save_setup,
1595     .save_live_iterate = htab_save_iterate,
1596     .save_live_complete_precopy = htab_save_complete,
1597     .load_state = htab_load,
1598 };
1599 
1600 static void spapr_boot_set(void *opaque, const char *boot_device,
1601                            Error **errp)
1602 {
1603     MachineState *machine = MACHINE(qdev_get_machine());
1604     machine->boot_order = g_strdup(boot_device);
1605 }
1606 
1607 static void spapr_cpu_init(sPAPRMachineState *spapr, PowerPCCPU *cpu,
1608                            Error **errp)
1609 {
1610     CPUPPCState *env = &cpu->env;
1611 
1612     /* Set time-base frequency to 512 MHz */
1613     cpu_ppc_tb_init(env, TIMEBASE_FREQ);
1614 
1615     /* PAPR always has exception vectors in RAM not ROM. To ensure this,
1616      * MSR[IP] should never be set.
1617      */
1618     env->msr_mask &= ~(1 << 6);
1619 
1620     /* Tell KVM that we're in PAPR mode */
1621     if (kvm_enabled()) {
1622         kvmppc_set_papr(cpu);
1623     }
1624 
1625     if (cpu->max_compat) {
1626         Error *local_err = NULL;
1627 
1628         ppc_set_compat(cpu, cpu->max_compat, &local_err);
1629         if (local_err) {
1630             error_propagate(errp, local_err);
1631             return;
1632         }
1633     }
1634 
1635     xics_cpu_setup(spapr->icp, cpu);
1636 
1637     qemu_register_reset(spapr_cpu_reset, cpu);
1638 }
1639 
1640 /*
1641  * Reset routine for LMB DR devices.
1642  *
1643  * Unlike PCI DR devices, LMB DR devices explicitly register this reset
1644  * routine. Reset for PCI DR devices will be handled by PHB reset routine
1645  * when it walks all its children devices. LMB devices reset occurs
1646  * as part of spapr_ppc_reset().
1647  */
1648 static void spapr_drc_reset(void *opaque)
1649 {
1650     sPAPRDRConnector *drc = opaque;
1651     DeviceState *d = DEVICE(drc);
1652 
1653     if (d) {
1654         device_reset(d);
1655     }
1656 }
1657 
1658 static void spapr_create_lmb_dr_connectors(sPAPRMachineState *spapr)
1659 {
1660     MachineState *machine = MACHINE(spapr);
1661     uint64_t lmb_size = SPAPR_MEMORY_BLOCK_SIZE;
1662     uint32_t nr_lmbs = (machine->maxram_size - machine->ram_size)/lmb_size;
1663     int i;
1664 
1665     for (i = 0; i < nr_lmbs; i++) {
1666         sPAPRDRConnector *drc;
1667         uint64_t addr;
1668 
1669         addr = i * lmb_size + spapr->hotplug_memory.base;
1670         drc = spapr_dr_connector_new(OBJECT(spapr), SPAPR_DR_CONNECTOR_TYPE_LMB,
1671                                      addr/lmb_size);
1672         qemu_register_reset(spapr_drc_reset, drc);
1673     }
1674 }
1675 
1676 /*
1677  * If RAM size, maxmem size and individual node mem sizes aren't aligned
1678  * to SPAPR_MEMORY_BLOCK_SIZE(256MB), then refuse to start the guest
1679  * since we can't support such unaligned sizes with DRCONF_MEMORY.
1680  */
1681 static void spapr_validate_node_memory(MachineState *machine, Error **errp)
1682 {
1683     int i;
1684 
1685     if (machine->ram_size % SPAPR_MEMORY_BLOCK_SIZE) {
1686         error_setg(errp, "Memory size 0x" RAM_ADDR_FMT
1687                    " is not aligned to %llu MiB",
1688                    machine->ram_size,
1689                    SPAPR_MEMORY_BLOCK_SIZE / M_BYTE);
1690         return;
1691     }
1692 
1693     if (machine->maxram_size % SPAPR_MEMORY_BLOCK_SIZE) {
1694         error_setg(errp, "Maximum memory size 0x" RAM_ADDR_FMT
1695                    " is not aligned to %llu MiB",
1696                    machine->ram_size,
1697                    SPAPR_MEMORY_BLOCK_SIZE / M_BYTE);
1698         return;
1699     }
1700 
1701     for (i = 0; i < nb_numa_nodes; i++) {
1702         if (numa_info[i].node_mem % SPAPR_MEMORY_BLOCK_SIZE) {
1703             error_setg(errp,
1704                        "Node %d memory size 0x%" PRIx64
1705                        " is not aligned to %llu MiB",
1706                        i, numa_info[i].node_mem,
1707                        SPAPR_MEMORY_BLOCK_SIZE / M_BYTE);
1708             return;
1709         }
1710     }
1711 }
1712 
1713 /* pSeries LPAR / sPAPR hardware init */
1714 static void ppc_spapr_init(MachineState *machine)
1715 {
1716     sPAPRMachineState *spapr = SPAPR_MACHINE(machine);
1717     sPAPRMachineClass *smc = SPAPR_MACHINE_GET_CLASS(machine);
1718     const char *kernel_filename = machine->kernel_filename;
1719     const char *kernel_cmdline = machine->kernel_cmdline;
1720     const char *initrd_filename = machine->initrd_filename;
1721     PowerPCCPU *cpu;
1722     PCIHostState *phb;
1723     int i;
1724     MemoryRegion *sysmem = get_system_memory();
1725     MemoryRegion *ram = g_new(MemoryRegion, 1);
1726     MemoryRegion *rma_region;
1727     void *rma = NULL;
1728     hwaddr rma_alloc_size;
1729     hwaddr node0_size = spapr_node0_size();
1730     uint32_t initrd_base = 0;
1731     long kernel_size = 0, initrd_size = 0;
1732     long load_limit, fw_size;
1733     bool kernel_le = false;
1734     char *filename;
1735 
1736     msi_nonbroken = true;
1737 
1738     QLIST_INIT(&spapr->phbs);
1739 
1740     cpu_ppc_hypercall = emulate_spapr_hypercall;
1741 
1742     /* Allocate RMA if necessary */
1743     rma_alloc_size = kvmppc_alloc_rma(&rma);
1744 
1745     if (rma_alloc_size == -1) {
1746         error_report("Unable to create RMA");
1747         exit(1);
1748     }
1749 
1750     if (rma_alloc_size && (rma_alloc_size < node0_size)) {
1751         spapr->rma_size = rma_alloc_size;
1752     } else {
1753         spapr->rma_size = node0_size;
1754 
1755         /* With KVM, we don't actually know whether KVM supports an
1756          * unbounded RMA (PR KVM) or is limited by the hash table size
1757          * (HV KVM using VRMA), so we always assume the latter
1758          *
1759          * In that case, we also limit the initial allocations for RTAS
1760          * etc... to 256M since we have no way to know what the VRMA size
1761          * is going to be as it depends on the size of the hash table
1762          * isn't determined yet.
1763          */
1764         if (kvm_enabled()) {
1765             spapr->vrma_adjust = 1;
1766             spapr->rma_size = MIN(spapr->rma_size, 0x10000000);
1767         }
1768     }
1769 
1770     if (spapr->rma_size > node0_size) {
1771         error_report("Numa node 0 has to span the RMA (%#08"HWADDR_PRIx")",
1772                      spapr->rma_size);
1773         exit(1);
1774     }
1775 
1776     /* Setup a load limit for the ramdisk leaving room for SLOF and FDT */
1777     load_limit = MIN(spapr->rma_size, RTAS_MAX_ADDR) - FW_OVERHEAD;
1778 
1779     /* Set up Interrupt Controller before we create the VCPUs */
1780     spapr->icp = xics_system_init(machine,
1781                                   DIV_ROUND_UP(max_cpus * kvmppc_smt_threads(),
1782                                                smp_threads),
1783                                   XICS_IRQS, &error_fatal);
1784 
1785     if (smc->dr_lmb_enabled) {
1786         spapr_validate_node_memory(machine, &error_fatal);
1787     }
1788 
1789     /* init CPUs */
1790     if (machine->cpu_model == NULL) {
1791         machine->cpu_model = kvm_enabled() ? "host" : "POWER7";
1792     }
1793     for (i = 0; i < smp_cpus; i++) {
1794         cpu = cpu_ppc_init(machine->cpu_model);
1795         if (cpu == NULL) {
1796             error_report("Unable to find PowerPC CPU definition");
1797             exit(1);
1798         }
1799         spapr_cpu_init(spapr, cpu, &error_fatal);
1800     }
1801 
1802     if (kvm_enabled()) {
1803         /* Enable H_LOGICAL_CI_* so SLOF can talk to in-kernel devices */
1804         kvmppc_enable_logical_ci_hcalls();
1805         kvmppc_enable_set_mode_hcall();
1806     }
1807 
1808     /* allocate RAM */
1809     memory_region_allocate_system_memory(ram, NULL, "ppc_spapr.ram",
1810                                          machine->ram_size);
1811     memory_region_add_subregion(sysmem, 0, ram);
1812 
1813     if (rma_alloc_size && rma) {
1814         rma_region = g_new(MemoryRegion, 1);
1815         memory_region_init_ram_ptr(rma_region, NULL, "ppc_spapr.rma",
1816                                    rma_alloc_size, rma);
1817         vmstate_register_ram_global(rma_region);
1818         memory_region_add_subregion(sysmem, 0, rma_region);
1819     }
1820 
1821     /* initialize hotplug memory address space */
1822     if (machine->ram_size < machine->maxram_size) {
1823         ram_addr_t hotplug_mem_size = machine->maxram_size - machine->ram_size;
1824 
1825         if (machine->ram_slots > SPAPR_MAX_RAM_SLOTS) {
1826             error_report("Specified number of memory slots %"
1827                          PRIu64" exceeds max supported %d",
1828                          machine->ram_slots, SPAPR_MAX_RAM_SLOTS);
1829             exit(1);
1830         }
1831 
1832         spapr->hotplug_memory.base = ROUND_UP(machine->ram_size,
1833                                               SPAPR_HOTPLUG_MEM_ALIGN);
1834         memory_region_init(&spapr->hotplug_memory.mr, OBJECT(spapr),
1835                            "hotplug-memory", hotplug_mem_size);
1836         memory_region_add_subregion(sysmem, spapr->hotplug_memory.base,
1837                                     &spapr->hotplug_memory.mr);
1838     }
1839 
1840     if (smc->dr_lmb_enabled) {
1841         spapr_create_lmb_dr_connectors(spapr);
1842     }
1843 
1844     filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, "spapr-rtas.bin");
1845     if (!filename) {
1846         error_report("Could not find LPAR rtas '%s'", "spapr-rtas.bin");
1847         exit(1);
1848     }
1849     spapr->rtas_size = get_image_size(filename);
1850     spapr->rtas_blob = g_malloc(spapr->rtas_size);
1851     if (load_image_size(filename, spapr->rtas_blob, spapr->rtas_size) < 0) {
1852         error_report("Could not load LPAR rtas '%s'", filename);
1853         exit(1);
1854     }
1855     if (spapr->rtas_size > RTAS_MAX_SIZE) {
1856         error_report("RTAS too big ! 0x%zx bytes (max is 0x%x)",
1857                      (size_t)spapr->rtas_size, RTAS_MAX_SIZE);
1858         exit(1);
1859     }
1860     g_free(filename);
1861 
1862     /* Set up EPOW events infrastructure */
1863     spapr_events_init(spapr);
1864 
1865     /* Set up the RTC RTAS interfaces */
1866     spapr_rtc_create(spapr);
1867 
1868     /* Set up VIO bus */
1869     spapr->vio_bus = spapr_vio_bus_init();
1870 
1871     for (i = 0; i < MAX_SERIAL_PORTS; i++) {
1872         if (serial_hds[i]) {
1873             spapr_vty_create(spapr->vio_bus, serial_hds[i]);
1874         }
1875     }
1876 
1877     /* We always have at least the nvram device on VIO */
1878     spapr_create_nvram(spapr);
1879 
1880     /* Set up PCI */
1881     spapr_pci_rtas_init();
1882 
1883     phb = spapr_create_phb(spapr, 0);
1884 
1885     for (i = 0; i < nb_nics; i++) {
1886         NICInfo *nd = &nd_table[i];
1887 
1888         if (!nd->model) {
1889             nd->model = g_strdup("ibmveth");
1890         }
1891 
1892         if (strcmp(nd->model, "ibmveth") == 0) {
1893             spapr_vlan_create(spapr->vio_bus, nd);
1894         } else {
1895             pci_nic_init_nofail(&nd_table[i], phb->bus, nd->model, NULL);
1896         }
1897     }
1898 
1899     for (i = 0; i <= drive_get_max_bus(IF_SCSI); i++) {
1900         spapr_vscsi_create(spapr->vio_bus);
1901     }
1902 
1903     /* Graphics */
1904     if (spapr_vga_init(phb->bus, &error_fatal)) {
1905         spapr->has_graphics = true;
1906         machine->usb |= defaults_enabled() && !machine->usb_disabled;
1907     }
1908 
1909     if (machine->usb) {
1910         if (smc->use_ohci_by_default) {
1911             pci_create_simple(phb->bus, -1, "pci-ohci");
1912         } else {
1913             pci_create_simple(phb->bus, -1, "nec-usb-xhci");
1914         }
1915 
1916         if (spapr->has_graphics) {
1917             USBBus *usb_bus = usb_bus_find(-1);
1918 
1919             usb_create_simple(usb_bus, "usb-kbd");
1920             usb_create_simple(usb_bus, "usb-mouse");
1921         }
1922     }
1923 
1924     if (spapr->rma_size < (MIN_RMA_SLOF << 20)) {
1925         error_report(
1926             "pSeries SLOF firmware requires >= %ldM guest RMA (Real Mode Area memory)",
1927             MIN_RMA_SLOF);
1928         exit(1);
1929     }
1930 
1931     if (kernel_filename) {
1932         uint64_t lowaddr = 0;
1933 
1934         kernel_size = load_elf(kernel_filename, translate_kernel_address, NULL,
1935                                NULL, &lowaddr, NULL, 1, PPC_ELF_MACHINE,
1936                                0, 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,
1941                                    0, 0);
1942             kernel_le = kernel_size > 0;
1943         }
1944         if (kernel_size < 0) {
1945             error_report("error loading %s: %s",
1946                          kernel_filename, load_elf_strerror(kernel_size));
1947             exit(1);
1948         }
1949 
1950         /* load initrd */
1951         if (initrd_filename) {
1952             /* Try to locate the initrd in the gap between the kernel
1953              * and the firmware. Add a bit of space just in case
1954              */
1955             initrd_base = (KERNEL_LOAD_ADDR + kernel_size + 0x1ffff) & ~0xffff;
1956             initrd_size = load_image_targphys(initrd_filename, initrd_base,
1957                                               load_limit - initrd_base);
1958             if (initrd_size < 0) {
1959                 error_report("could not load initial ram disk '%s'",
1960                              initrd_filename);
1961                 exit(1);
1962             }
1963         } else {
1964             initrd_base = 0;
1965             initrd_size = 0;
1966         }
1967     }
1968 
1969     if (bios_name == NULL) {
1970         bios_name = FW_FILE_NAME;
1971     }
1972     filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, bios_name);
1973     if (!filename) {
1974         error_report("Could not find LPAR firmware '%s'", bios_name);
1975         exit(1);
1976     }
1977     fw_size = load_image_targphys(filename, 0, FW_MAX_SIZE);
1978     if (fw_size <= 0) {
1979         error_report("Could not load LPAR firmware '%s'", filename);
1980         exit(1);
1981     }
1982     g_free(filename);
1983 
1984     /* FIXME: Should register things through the MachineState's qdev
1985      * interface, this is a legacy from the sPAPREnvironment structure
1986      * which predated MachineState but had a similar function */
1987     vmstate_register(NULL, 0, &vmstate_spapr, spapr);
1988     register_savevm_live(NULL, "spapr/htab", -1, 1,
1989                          &savevm_htab_handlers, spapr);
1990 
1991     /* Prepare the device tree */
1992     spapr->fdt_skel = spapr_create_fdt_skel(initrd_base, initrd_size,
1993                                             kernel_size, kernel_le,
1994                                             kernel_cmdline,
1995                                             spapr->check_exception_irq);
1996     assert(spapr->fdt_skel != NULL);
1997 
1998     /* used by RTAS */
1999     QTAILQ_INIT(&spapr->ccs_list);
2000     qemu_register_reset(spapr_ccs_reset_hook, spapr);
2001 
2002     qemu_register_boot_set(spapr_boot_set, spapr);
2003 }
2004 
2005 static int spapr_kvm_type(const char *vm_type)
2006 {
2007     if (!vm_type) {
2008         return 0;
2009     }
2010 
2011     if (!strcmp(vm_type, "HV")) {
2012         return 1;
2013     }
2014 
2015     if (!strcmp(vm_type, "PR")) {
2016         return 2;
2017     }
2018 
2019     error_report("Unknown kvm-type specified '%s'", vm_type);
2020     exit(1);
2021 }
2022 
2023 /*
2024  * Implementation of an interface to adjust firmware path
2025  * for the bootindex property handling.
2026  */
2027 static char *spapr_get_fw_dev_path(FWPathProvider *p, BusState *bus,
2028                                    DeviceState *dev)
2029 {
2030 #define CAST(type, obj, name) \
2031     ((type *)object_dynamic_cast(OBJECT(obj), (name)))
2032     SCSIDevice *d = CAST(SCSIDevice,  dev, TYPE_SCSI_DEVICE);
2033     sPAPRPHBState *phb = CAST(sPAPRPHBState, dev, TYPE_SPAPR_PCI_HOST_BRIDGE);
2034 
2035     if (d) {
2036         void *spapr = CAST(void, bus->parent, "spapr-vscsi");
2037         VirtIOSCSI *virtio = CAST(VirtIOSCSI, bus->parent, TYPE_VIRTIO_SCSI);
2038         USBDevice *usb = CAST(USBDevice, bus->parent, TYPE_USB_DEVICE);
2039 
2040         if (spapr) {
2041             /*
2042              * Replace "channel@0/disk@0,0" with "disk@8000000000000000":
2043              * We use SRP luns of the form 8000 | (bus << 8) | (id << 5) | lun
2044              * in the top 16 bits of the 64-bit LUN
2045              */
2046             unsigned id = 0x8000 | (d->id << 8) | d->lun;
2047             return g_strdup_printf("%s@%"PRIX64, qdev_fw_name(dev),
2048                                    (uint64_t)id << 48);
2049         } else if (virtio) {
2050             /*
2051              * We use SRP luns of the form 01000000 | (target << 8) | lun
2052              * in the top 32 bits of the 64-bit LUN
2053              * Note: the quote above is from SLOF and it is wrong,
2054              * the actual binding is:
2055              * swap 0100 or 10 << or 20 << ( target lun-id -- srplun )
2056              */
2057             unsigned id = 0x1000000 | (d->id << 16) | d->lun;
2058             return g_strdup_printf("%s@%"PRIX64, qdev_fw_name(dev),
2059                                    (uint64_t)id << 32);
2060         } else if (usb) {
2061             /*
2062              * We use SRP luns of the form 01000000 | (usb-port << 16) | lun
2063              * in the top 32 bits of the 64-bit LUN
2064              */
2065             unsigned usb_port = atoi(usb->port->path);
2066             unsigned id = 0x1000000 | (usb_port << 16) | d->lun;
2067             return g_strdup_printf("%s@%"PRIX64, qdev_fw_name(dev),
2068                                    (uint64_t)id << 32);
2069         }
2070     }
2071 
2072     if (phb) {
2073         /* Replace "pci" with "pci@800000020000000" */
2074         return g_strdup_printf("pci@%"PRIX64, phb->buid);
2075     }
2076 
2077     return NULL;
2078 }
2079 
2080 static char *spapr_get_kvm_type(Object *obj, Error **errp)
2081 {
2082     sPAPRMachineState *spapr = SPAPR_MACHINE(obj);
2083 
2084     return g_strdup(spapr->kvm_type);
2085 }
2086 
2087 static void spapr_set_kvm_type(Object *obj, const char *value, Error **errp)
2088 {
2089     sPAPRMachineState *spapr = SPAPR_MACHINE(obj);
2090 
2091     g_free(spapr->kvm_type);
2092     spapr->kvm_type = g_strdup(value);
2093 }
2094 
2095 static void spapr_machine_initfn(Object *obj)
2096 {
2097     sPAPRMachineState *spapr = SPAPR_MACHINE(obj);
2098 
2099     spapr->htab_fd = -1;
2100     object_property_add_str(obj, "kvm-type",
2101                             spapr_get_kvm_type, spapr_set_kvm_type, NULL);
2102     object_property_set_description(obj, "kvm-type",
2103                                     "Specifies the KVM virtualization mode (HV, PR)",
2104                                     NULL);
2105 }
2106 
2107 static void spapr_machine_finalizefn(Object *obj)
2108 {
2109     sPAPRMachineState *spapr = SPAPR_MACHINE(obj);
2110 
2111     g_free(spapr->kvm_type);
2112 }
2113 
2114 static void ppc_cpu_do_nmi_on_cpu(void *arg)
2115 {
2116     CPUState *cs = arg;
2117 
2118     cpu_synchronize_state(cs);
2119     ppc_cpu_do_system_reset(cs);
2120 }
2121 
2122 static void spapr_nmi(NMIState *n, int cpu_index, Error **errp)
2123 {
2124     CPUState *cs;
2125 
2126     CPU_FOREACH(cs) {
2127         async_run_on_cpu(cs, ppc_cpu_do_nmi_on_cpu, cs);
2128     }
2129 }
2130 
2131 static void spapr_add_lmbs(DeviceState *dev, uint64_t addr, uint64_t size,
2132                            uint32_t node, Error **errp)
2133 {
2134     sPAPRDRConnector *drc;
2135     sPAPRDRConnectorClass *drck;
2136     uint32_t nr_lmbs = size/SPAPR_MEMORY_BLOCK_SIZE;
2137     int i, fdt_offset, fdt_size;
2138     void *fdt;
2139 
2140     /*
2141      * Check for DRC connectors and send hotplug notification to the
2142      * guest only in case of hotplugged memory. This allows cold plugged
2143      * memory to be specified at boot time.
2144      */
2145     if (!dev->hotplugged) {
2146         return;
2147     }
2148 
2149     for (i = 0; i < nr_lmbs; i++) {
2150         drc = spapr_dr_connector_by_id(SPAPR_DR_CONNECTOR_TYPE_LMB,
2151                 addr/SPAPR_MEMORY_BLOCK_SIZE);
2152         g_assert(drc);
2153 
2154         fdt = create_device_tree(&fdt_size);
2155         fdt_offset = spapr_populate_memory_node(fdt, node, addr,
2156                                                 SPAPR_MEMORY_BLOCK_SIZE);
2157 
2158         drck = SPAPR_DR_CONNECTOR_GET_CLASS(drc);
2159         drck->attach(drc, dev, fdt, fdt_offset, !dev->hotplugged, errp);
2160         addr += SPAPR_MEMORY_BLOCK_SIZE;
2161     }
2162     spapr_hotplug_req_add_by_count(SPAPR_DR_CONNECTOR_TYPE_LMB, nr_lmbs);
2163 }
2164 
2165 static void spapr_memory_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
2166                               uint32_t node, Error **errp)
2167 {
2168     Error *local_err = NULL;
2169     sPAPRMachineState *ms = SPAPR_MACHINE(hotplug_dev);
2170     PCDIMMDevice *dimm = PC_DIMM(dev);
2171     PCDIMMDeviceClass *ddc = PC_DIMM_GET_CLASS(dimm);
2172     MemoryRegion *mr = ddc->get_memory_region(dimm);
2173     uint64_t align = memory_region_get_alignment(mr);
2174     uint64_t size = memory_region_size(mr);
2175     uint64_t addr;
2176 
2177     if (size % SPAPR_MEMORY_BLOCK_SIZE) {
2178         error_setg(&local_err, "Hotplugged memory size must be a multiple of "
2179                       "%lld MB", SPAPR_MEMORY_BLOCK_SIZE/M_BYTE);
2180         goto out;
2181     }
2182 
2183     pc_dimm_memory_plug(dev, &ms->hotplug_memory, mr, align, &local_err);
2184     if (local_err) {
2185         goto out;
2186     }
2187 
2188     addr = object_property_get_int(OBJECT(dimm), PC_DIMM_ADDR_PROP, &local_err);
2189     if (local_err) {
2190         pc_dimm_memory_unplug(dev, &ms->hotplug_memory, mr);
2191         goto out;
2192     }
2193 
2194     spapr_add_lmbs(dev, addr, size, node, &error_abort);
2195 
2196 out:
2197     error_propagate(errp, local_err);
2198 }
2199 
2200 static void spapr_machine_device_plug(HotplugHandler *hotplug_dev,
2201                                       DeviceState *dev, Error **errp)
2202 {
2203     sPAPRMachineClass *smc = SPAPR_MACHINE_GET_CLASS(qdev_get_machine());
2204 
2205     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2206         int node;
2207 
2208         if (!smc->dr_lmb_enabled) {
2209             error_setg(errp, "Memory hotplug not supported for this machine");
2210             return;
2211         }
2212         node = object_property_get_int(OBJECT(dev), PC_DIMM_NODE_PROP, errp);
2213         if (*errp) {
2214             return;
2215         }
2216 
2217         /*
2218          * Currently PowerPC kernel doesn't allow hot-adding memory to
2219          * memory-less node, but instead will silently add the memory
2220          * to the first node that has some memory. This causes two
2221          * unexpected behaviours for the user.
2222          *
2223          * - Memory gets hotplugged to a different node than what the user
2224          *   specified.
2225          * - Since pc-dimm subsystem in QEMU still thinks that memory belongs
2226          *   to memory-less node, a reboot will set things accordingly
2227          *   and the previously hotplugged memory now ends in the right node.
2228          *   This appears as if some memory moved from one node to another.
2229          *
2230          * So until kernel starts supporting memory hotplug to memory-less
2231          * nodes, just prevent such attempts upfront in QEMU.
2232          */
2233         if (nb_numa_nodes && !numa_info[node].node_mem) {
2234             error_setg(errp, "Can't hotplug memory to memory-less node %d",
2235                        node);
2236             return;
2237         }
2238 
2239         spapr_memory_plug(hotplug_dev, dev, node, errp);
2240     }
2241 }
2242 
2243 static void spapr_machine_device_unplug(HotplugHandler *hotplug_dev,
2244                                       DeviceState *dev, Error **errp)
2245 {
2246     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2247         error_setg(errp, "Memory hot unplug not supported by sPAPR");
2248     }
2249 }
2250 
2251 static HotplugHandler *spapr_get_hotpug_handler(MachineState *machine,
2252                                              DeviceState *dev)
2253 {
2254     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2255         return HOTPLUG_HANDLER(machine);
2256     }
2257     return NULL;
2258 }
2259 
2260 static unsigned spapr_cpu_index_to_socket_id(unsigned cpu_index)
2261 {
2262     /* Allocate to NUMA nodes on a "socket" basis (not that concept of
2263      * socket means much for the paravirtualized PAPR platform) */
2264     return cpu_index / smp_threads / smp_cores;
2265 }
2266 
2267 static void spapr_machine_class_init(ObjectClass *oc, void *data)
2268 {
2269     MachineClass *mc = MACHINE_CLASS(oc);
2270     sPAPRMachineClass *smc = SPAPR_MACHINE_CLASS(oc);
2271     FWPathProviderClass *fwc = FW_PATH_PROVIDER_CLASS(oc);
2272     NMIClass *nc = NMI_CLASS(oc);
2273     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
2274 
2275     mc->desc = "pSeries Logical Partition (PAPR compliant)";
2276 
2277     /*
2278      * We set up the default / latest behaviour here.  The class_init
2279      * functions for the specific versioned machine types can override
2280      * these details for backwards compatibility
2281      */
2282     mc->init = ppc_spapr_init;
2283     mc->reset = ppc_spapr_reset;
2284     mc->block_default_type = IF_SCSI;
2285     mc->max_cpus = MAX_CPUMASK_BITS;
2286     mc->no_parallel = 1;
2287     mc->default_boot_order = "";
2288     mc->default_ram_size = 512 * M_BYTE;
2289     mc->kvm_type = spapr_kvm_type;
2290     mc->has_dynamic_sysbus = true;
2291     mc->pci_allow_0_address = true;
2292     mc->get_hotplug_handler = spapr_get_hotpug_handler;
2293     hc->plug = spapr_machine_device_plug;
2294     hc->unplug = spapr_machine_device_unplug;
2295     mc->cpu_index_to_socket_id = spapr_cpu_index_to_socket_id;
2296 
2297     smc->dr_lmb_enabled = true;
2298     fwc->get_dev_path = spapr_get_fw_dev_path;
2299     nc->nmi_monitor_handler = spapr_nmi;
2300 }
2301 
2302 static const TypeInfo spapr_machine_info = {
2303     .name          = TYPE_SPAPR_MACHINE,
2304     .parent        = TYPE_MACHINE,
2305     .abstract      = true,
2306     .instance_size = sizeof(sPAPRMachineState),
2307     .instance_init = spapr_machine_initfn,
2308     .instance_finalize = spapr_machine_finalizefn,
2309     .class_size    = sizeof(sPAPRMachineClass),
2310     .class_init    = spapr_machine_class_init,
2311     .interfaces = (InterfaceInfo[]) {
2312         { TYPE_FW_PATH_PROVIDER },
2313         { TYPE_NMI },
2314         { TYPE_HOTPLUG_HANDLER },
2315         { }
2316     },
2317 };
2318 
2319 #define DEFINE_SPAPR_MACHINE(suffix, verstr, latest)                 \
2320     static void spapr_machine_##suffix##_class_init(ObjectClass *oc, \
2321                                                     void *data)      \
2322     {                                                                \
2323         MachineClass *mc = MACHINE_CLASS(oc);                        \
2324         spapr_machine_##suffix##_class_options(mc);                  \
2325         if (latest) {                                                \
2326             mc->alias = "pseries";                                   \
2327             mc->is_default = 1;                                      \
2328         }                                                            \
2329     }                                                                \
2330     static void spapr_machine_##suffix##_instance_init(Object *obj)  \
2331     {                                                                \
2332         MachineState *machine = MACHINE(obj);                        \
2333         spapr_machine_##suffix##_instance_options(machine);          \
2334     }                                                                \
2335     static const TypeInfo spapr_machine_##suffix##_info = {          \
2336         .name = MACHINE_TYPE_NAME("pseries-" verstr),                \
2337         .parent = TYPE_SPAPR_MACHINE,                                \
2338         .class_init = spapr_machine_##suffix##_class_init,           \
2339         .instance_init = spapr_machine_##suffix##_instance_init,     \
2340     };                                                               \
2341     static void spapr_machine_register_##suffix(void)                \
2342     {                                                                \
2343         type_register(&spapr_machine_##suffix##_info);               \
2344     }                                                                \
2345     type_init(spapr_machine_register_##suffix)
2346 
2347 /*
2348  * pseries-2.6
2349  */
2350 static void spapr_machine_2_6_instance_options(MachineState *machine)
2351 {
2352 }
2353 
2354 static void spapr_machine_2_6_class_options(MachineClass *mc)
2355 {
2356     /* Defaults for the latest behaviour inherited from the base class */
2357 }
2358 
2359 DEFINE_SPAPR_MACHINE(2_6, "2.6", true);
2360 
2361 /*
2362  * pseries-2.5
2363  */
2364 #define SPAPR_COMPAT_2_5 \
2365         HW_COMPAT_2_5
2366 
2367 static void spapr_machine_2_5_instance_options(MachineState *machine)
2368 {
2369 }
2370 
2371 static void spapr_machine_2_5_class_options(MachineClass *mc)
2372 {
2373     sPAPRMachineClass *smc = SPAPR_MACHINE_CLASS(mc);
2374 
2375     spapr_machine_2_6_class_options(mc);
2376     smc->use_ohci_by_default = true;
2377     SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_5);
2378 }
2379 
2380 DEFINE_SPAPR_MACHINE(2_5, "2.5", false);
2381 
2382 /*
2383  * pseries-2.4
2384  */
2385 #define SPAPR_COMPAT_2_4 \
2386         SPAPR_COMPAT_2_5 \
2387         HW_COMPAT_2_4
2388 
2389 static void spapr_machine_2_4_instance_options(MachineState *machine)
2390 {
2391     spapr_machine_2_5_instance_options(machine);
2392 }
2393 
2394 static void spapr_machine_2_4_class_options(MachineClass *mc)
2395 {
2396     sPAPRMachineClass *smc = SPAPR_MACHINE_CLASS(mc);
2397 
2398     spapr_machine_2_5_class_options(mc);
2399     smc->dr_lmb_enabled = false;
2400     SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_4);
2401 }
2402 
2403 DEFINE_SPAPR_MACHINE(2_4, "2.4", false);
2404 
2405 /*
2406  * pseries-2.3
2407  */
2408 #define SPAPR_COMPAT_2_3 \
2409         SPAPR_COMPAT_2_4 \
2410         HW_COMPAT_2_3 \
2411         {\
2412             .driver   = "spapr-pci-host-bridge",\
2413             .property = "dynamic-reconfiguration",\
2414             .value    = "off",\
2415         },
2416 
2417 static void spapr_machine_2_3_instance_options(MachineState *machine)
2418 {
2419     spapr_machine_2_4_instance_options(machine);
2420     savevm_skip_section_footers();
2421     global_state_set_optional();
2422     savevm_skip_configuration();
2423 }
2424 
2425 static void spapr_machine_2_3_class_options(MachineClass *mc)
2426 {
2427     spapr_machine_2_4_class_options(mc);
2428     SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_3);
2429 }
2430 DEFINE_SPAPR_MACHINE(2_3, "2.3", false);
2431 
2432 /*
2433  * pseries-2.2
2434  */
2435 
2436 #define SPAPR_COMPAT_2_2 \
2437         SPAPR_COMPAT_2_3 \
2438         HW_COMPAT_2_2 \
2439         {\
2440             .driver   = TYPE_SPAPR_PCI_HOST_BRIDGE,\
2441             .property = "mem_win_size",\
2442             .value    = "0x20000000",\
2443         },
2444 
2445 static void spapr_machine_2_2_instance_options(MachineState *machine)
2446 {
2447     spapr_machine_2_3_instance_options(machine);
2448     machine->suppress_vmdesc = true;
2449 }
2450 
2451 static void spapr_machine_2_2_class_options(MachineClass *mc)
2452 {
2453     spapr_machine_2_3_class_options(mc);
2454     SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_2);
2455 }
2456 DEFINE_SPAPR_MACHINE(2_2, "2.2", false);
2457 
2458 /*
2459  * pseries-2.1
2460  */
2461 #define SPAPR_COMPAT_2_1 \
2462         SPAPR_COMPAT_2_2 \
2463         HW_COMPAT_2_1
2464 
2465 static void spapr_machine_2_1_instance_options(MachineState *machine)
2466 {
2467     spapr_machine_2_2_instance_options(machine);
2468 }
2469 
2470 static void spapr_machine_2_1_class_options(MachineClass *mc)
2471 {
2472     spapr_machine_2_2_class_options(mc);
2473     SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_1);
2474 }
2475 DEFINE_SPAPR_MACHINE(2_1, "2.1", false);
2476 
2477 static void spapr_machine_register_types(void)
2478 {
2479     type_register_static(&spapr_machine_info);
2480 }
2481 
2482 type_init(spapr_machine_register_types)
2483