xref: /openbmc/qemu/hw/ppc/spapr.c (revision d341d9f3)
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_reportf_err(err,
126                               "kernel_irqchip requested but unavailable: ");
127         } else {
128             error_free(err);
129         }
130     }
131 
132     if (!icp) {
133         icp = try_create_xics(TYPE_XICS, nr_servers, nr_irqs, &error_abort);
134     }
135 
136     return icp;
137 }
138 
139 static int spapr_fixup_cpu_smt_dt(void *fdt, int offset, PowerPCCPU *cpu,
140                                   int smt_threads)
141 {
142     int i, ret = 0;
143     uint32_t servers_prop[smt_threads];
144     uint32_t gservers_prop[smt_threads * 2];
145     int index = ppc_get_vcpu_dt_id(cpu);
146 
147     if (cpu->cpu_version) {
148         ret = fdt_setprop_cell(fdt, offset, "cpu-version", cpu->cpu_version);
149         if (ret < 0) {
150             return ret;
151         }
152     }
153 
154     /* Build interrupt servers and gservers properties */
155     for (i = 0; i < smt_threads; i++) {
156         servers_prop[i] = cpu_to_be32(index + i);
157         /* Hack, direct the group queues back to cpu 0 */
158         gservers_prop[i*2] = cpu_to_be32(index + i);
159         gservers_prop[i*2 + 1] = 0;
160     }
161     ret = fdt_setprop(fdt, offset, "ibm,ppc-interrupt-server#s",
162                       servers_prop, sizeof(servers_prop));
163     if (ret < 0) {
164         return ret;
165     }
166     ret = fdt_setprop(fdt, offset, "ibm,ppc-interrupt-gserver#s",
167                       gservers_prop, sizeof(gservers_prop));
168 
169     return ret;
170 }
171 
172 static int spapr_fixup_cpu_numa_dt(void *fdt, int offset, CPUState *cs)
173 {
174     int ret = 0;
175     PowerPCCPU *cpu = POWERPC_CPU(cs);
176     int index = ppc_get_vcpu_dt_id(cpu);
177     uint32_t associativity[] = {cpu_to_be32(0x5),
178                                 cpu_to_be32(0x0),
179                                 cpu_to_be32(0x0),
180                                 cpu_to_be32(0x0),
181                                 cpu_to_be32(cs->numa_node),
182                                 cpu_to_be32(index)};
183 
184     /* Advertise NUMA via ibm,associativity */
185     if (nb_numa_nodes > 1) {
186         ret = fdt_setprop(fdt, offset, "ibm,associativity", associativity,
187                           sizeof(associativity));
188     }
189 
190     return ret;
191 }
192 
193 static int spapr_fixup_cpu_dt(void *fdt, sPAPRMachineState *spapr)
194 {
195     int ret = 0, offset, cpus_offset;
196     CPUState *cs;
197     char cpu_model[32];
198     int smt = kvmppc_smt_threads();
199     uint32_t pft_size_prop[] = {0, cpu_to_be32(spapr->htab_shift)};
200 
201     CPU_FOREACH(cs) {
202         PowerPCCPU *cpu = POWERPC_CPU(cs);
203         DeviceClass *dc = DEVICE_GET_CLASS(cs);
204         int index = ppc_get_vcpu_dt_id(cpu);
205 
206         if ((index % smt) != 0) {
207             continue;
208         }
209 
210         snprintf(cpu_model, 32, "%s@%x", dc->fw_name, index);
211 
212         cpus_offset = fdt_path_offset(fdt, "/cpus");
213         if (cpus_offset < 0) {
214             cpus_offset = fdt_add_subnode(fdt, fdt_path_offset(fdt, "/"),
215                                           "cpus");
216             if (cpus_offset < 0) {
217                 return cpus_offset;
218             }
219         }
220         offset = fdt_subnode_offset(fdt, cpus_offset, cpu_model);
221         if (offset < 0) {
222             offset = fdt_add_subnode(fdt, cpus_offset, cpu_model);
223             if (offset < 0) {
224                 return offset;
225             }
226         }
227 
228         ret = fdt_setprop(fdt, offset, "ibm,pft-size",
229                           pft_size_prop, sizeof(pft_size_prop));
230         if (ret < 0) {
231             return ret;
232         }
233 
234         ret = spapr_fixup_cpu_numa_dt(fdt, offset, cs);
235         if (ret < 0) {
236             return ret;
237         }
238 
239         ret = spapr_fixup_cpu_smt_dt(fdt, offset, cpu,
240                                      ppc_get_compat_smt_threads(cpu));
241         if (ret < 0) {
242             return ret;
243         }
244     }
245     return ret;
246 }
247 
248 
249 static size_t create_page_sizes_prop(CPUPPCState *env, uint32_t *prop,
250                                      size_t maxsize)
251 {
252     size_t maxcells = maxsize / sizeof(uint32_t);
253     int i, j, count;
254     uint32_t *p = prop;
255 
256     for (i = 0; i < PPC_PAGE_SIZES_MAX_SZ; i++) {
257         struct ppc_one_seg_page_size *sps = &env->sps.sps[i];
258 
259         if (!sps->page_shift) {
260             break;
261         }
262         for (count = 0; count < PPC_PAGE_SIZES_MAX_SZ; count++) {
263             if (sps->enc[count].page_shift == 0) {
264                 break;
265             }
266         }
267         if ((p - prop) >= (maxcells - 3 - count * 2)) {
268             break;
269         }
270         *(p++) = cpu_to_be32(sps->page_shift);
271         *(p++) = cpu_to_be32(sps->slb_enc);
272         *(p++) = cpu_to_be32(count);
273         for (j = 0; j < count; j++) {
274             *(p++) = cpu_to_be32(sps->enc[j].page_shift);
275             *(p++) = cpu_to_be32(sps->enc[j].pte_enc);
276         }
277     }
278 
279     return (p - prop) * sizeof(uint32_t);
280 }
281 
282 static hwaddr spapr_node0_size(void)
283 {
284     MachineState *machine = MACHINE(qdev_get_machine());
285 
286     if (nb_numa_nodes) {
287         int i;
288         for (i = 0; i < nb_numa_nodes; ++i) {
289             if (numa_info[i].node_mem) {
290                 return MIN(pow2floor(numa_info[i].node_mem),
291                            machine->ram_size);
292             }
293         }
294     }
295     return machine->ram_size;
296 }
297 
298 #define _FDT(exp) \
299     do { \
300         int ret = (exp);                                           \
301         if (ret < 0) {                                             \
302             fprintf(stderr, "qemu: error creating device tree: %s: %s\n", \
303                     #exp, fdt_strerror(ret));                      \
304             exit(1);                                               \
305         }                                                          \
306     } while (0)
307 
308 static void add_str(GString *s, const gchar *s1)
309 {
310     g_string_append_len(s, s1, strlen(s1) + 1);
311 }
312 
313 static void *spapr_create_fdt_skel(hwaddr initrd_base,
314                                    hwaddr initrd_size,
315                                    hwaddr kernel_size,
316                                    bool little_endian,
317                                    const char *kernel_cmdline,
318                                    uint32_t epow_irq)
319 {
320     void *fdt;
321     uint32_t start_prop = cpu_to_be32(initrd_base);
322     uint32_t end_prop = cpu_to_be32(initrd_base + initrd_size);
323     GString *hypertas = g_string_sized_new(256);
324     GString *qemu_hypertas = g_string_sized_new(256);
325     uint32_t refpoints[] = {cpu_to_be32(0x4), cpu_to_be32(0x4)};
326     uint32_t interrupt_server_ranges_prop[] = {0, cpu_to_be32(max_cpus)};
327     unsigned char vec5[] = {0x0, 0x0, 0x0, 0x0, 0x0, 0x80};
328     char *buf;
329 
330     add_str(hypertas, "hcall-pft");
331     add_str(hypertas, "hcall-term");
332     add_str(hypertas, "hcall-dabr");
333     add_str(hypertas, "hcall-interrupt");
334     add_str(hypertas, "hcall-tce");
335     add_str(hypertas, "hcall-vio");
336     add_str(hypertas, "hcall-splpar");
337     add_str(hypertas, "hcall-bulk");
338     add_str(hypertas, "hcall-set-mode");
339     add_str(qemu_hypertas, "hcall-memop1");
340 
341     fdt = g_malloc0(FDT_MAX_SIZE);
342     _FDT((fdt_create(fdt, FDT_MAX_SIZE)));
343 
344     if (kernel_size) {
345         _FDT((fdt_add_reservemap_entry(fdt, KERNEL_LOAD_ADDR, kernel_size)));
346     }
347     if (initrd_size) {
348         _FDT((fdt_add_reservemap_entry(fdt, initrd_base, initrd_size)));
349     }
350     _FDT((fdt_finish_reservemap(fdt)));
351 
352     /* Root node */
353     _FDT((fdt_begin_node(fdt, "")));
354     _FDT((fdt_property_string(fdt, "device_type", "chrp")));
355     _FDT((fdt_property_string(fdt, "model", "IBM pSeries (emulated by qemu)")));
356     _FDT((fdt_property_string(fdt, "compatible", "qemu,pseries")));
357 
358     /*
359      * Add info to guest to indentify which host is it being run on
360      * and what is the uuid of the guest
361      */
362     if (kvmppc_get_host_model(&buf)) {
363         _FDT((fdt_property_string(fdt, "host-model", buf)));
364         g_free(buf);
365     }
366     if (kvmppc_get_host_serial(&buf)) {
367         _FDT((fdt_property_string(fdt, "host-serial", buf)));
368         g_free(buf);
369     }
370 
371     buf = g_strdup_printf(UUID_FMT, qemu_uuid[0], qemu_uuid[1],
372                           qemu_uuid[2], qemu_uuid[3], qemu_uuid[4],
373                           qemu_uuid[5], qemu_uuid[6], qemu_uuid[7],
374                           qemu_uuid[8], qemu_uuid[9], qemu_uuid[10],
375                           qemu_uuid[11], qemu_uuid[12], qemu_uuid[13],
376                           qemu_uuid[14], qemu_uuid[15]);
377 
378     _FDT((fdt_property_string(fdt, "vm,uuid", buf)));
379     if (qemu_uuid_set) {
380         _FDT((fdt_property_string(fdt, "system-id", buf)));
381     }
382     g_free(buf);
383 
384     if (qemu_get_vm_name()) {
385         _FDT((fdt_property_string(fdt, "ibm,partition-name",
386                                   qemu_get_vm_name())));
387     }
388 
389     _FDT((fdt_property_cell(fdt, "#address-cells", 0x2)));
390     _FDT((fdt_property_cell(fdt, "#size-cells", 0x2)));
391 
392     /* /chosen */
393     _FDT((fdt_begin_node(fdt, "chosen")));
394 
395     /* Set Form1_affinity */
396     _FDT((fdt_property(fdt, "ibm,architecture-vec-5", vec5, sizeof(vec5))));
397 
398     _FDT((fdt_property_string(fdt, "bootargs", kernel_cmdline)));
399     _FDT((fdt_property(fdt, "linux,initrd-start",
400                        &start_prop, sizeof(start_prop))));
401     _FDT((fdt_property(fdt, "linux,initrd-end",
402                        &end_prop, sizeof(end_prop))));
403     if (kernel_size) {
404         uint64_t kprop[2] = { cpu_to_be64(KERNEL_LOAD_ADDR),
405                               cpu_to_be64(kernel_size) };
406 
407         _FDT((fdt_property(fdt, "qemu,boot-kernel", &kprop, sizeof(kprop))));
408         if (little_endian) {
409             _FDT((fdt_property(fdt, "qemu,boot-kernel-le", NULL, 0)));
410         }
411     }
412     if (boot_menu) {
413         _FDT((fdt_property_cell(fdt, "qemu,boot-menu", boot_menu)));
414     }
415     _FDT((fdt_property_cell(fdt, "qemu,graphic-width", graphic_width)));
416     _FDT((fdt_property_cell(fdt, "qemu,graphic-height", graphic_height)));
417     _FDT((fdt_property_cell(fdt, "qemu,graphic-depth", graphic_depth)));
418 
419     _FDT((fdt_end_node(fdt)));
420 
421     /* RTAS */
422     _FDT((fdt_begin_node(fdt, "rtas")));
423 
424     if (!kvm_enabled() || kvmppc_spapr_use_multitce()) {
425         add_str(hypertas, "hcall-multi-tce");
426     }
427     _FDT((fdt_property(fdt, "ibm,hypertas-functions", hypertas->str,
428                        hypertas->len)));
429     g_string_free(hypertas, TRUE);
430     _FDT((fdt_property(fdt, "qemu,hypertas-functions", qemu_hypertas->str,
431                        qemu_hypertas->len)));
432     g_string_free(qemu_hypertas, TRUE);
433 
434     _FDT((fdt_property(fdt, "ibm,associativity-reference-points",
435         refpoints, sizeof(refpoints))));
436 
437     _FDT((fdt_property_cell(fdt, "rtas-error-log-max", RTAS_ERROR_LOG_MAX)));
438     _FDT((fdt_property_cell(fdt, "rtas-event-scan-rate",
439                             RTAS_EVENT_SCAN_RATE)));
440 
441     if (msi_supported) {
442         _FDT((fdt_property(fdt, "ibm,change-msix-capable", NULL, 0)));
443     }
444 
445     /*
446      * According to PAPR, rtas ibm,os-term does not guarantee a return
447      * back to the guest cpu.
448      *
449      * While an additional ibm,extended-os-term property indicates that
450      * rtas call return will always occur. Set this property.
451      */
452     _FDT((fdt_property(fdt, "ibm,extended-os-term", NULL, 0)));
453 
454     _FDT((fdt_end_node(fdt)));
455 
456     /* interrupt controller */
457     _FDT((fdt_begin_node(fdt, "interrupt-controller")));
458 
459     _FDT((fdt_property_string(fdt, "device_type",
460                               "PowerPC-External-Interrupt-Presentation")));
461     _FDT((fdt_property_string(fdt, "compatible", "IBM,ppc-xicp")));
462     _FDT((fdt_property(fdt, "interrupt-controller", NULL, 0)));
463     _FDT((fdt_property(fdt, "ibm,interrupt-server-ranges",
464                        interrupt_server_ranges_prop,
465                        sizeof(interrupt_server_ranges_prop))));
466     _FDT((fdt_property_cell(fdt, "#interrupt-cells", 2)));
467     _FDT((fdt_property_cell(fdt, "linux,phandle", PHANDLE_XICP)));
468     _FDT((fdt_property_cell(fdt, "phandle", PHANDLE_XICP)));
469 
470     _FDT((fdt_end_node(fdt)));
471 
472     /* vdevice */
473     _FDT((fdt_begin_node(fdt, "vdevice")));
474 
475     _FDT((fdt_property_string(fdt, "device_type", "vdevice")));
476     _FDT((fdt_property_string(fdt, "compatible", "IBM,vdevice")));
477     _FDT((fdt_property_cell(fdt, "#address-cells", 0x1)));
478     _FDT((fdt_property_cell(fdt, "#size-cells", 0x0)));
479     _FDT((fdt_property_cell(fdt, "#interrupt-cells", 0x2)));
480     _FDT((fdt_property(fdt, "interrupt-controller", NULL, 0)));
481 
482     _FDT((fdt_end_node(fdt)));
483 
484     /* event-sources */
485     spapr_events_fdt_skel(fdt, epow_irq);
486 
487     /* /hypervisor node */
488     if (kvm_enabled()) {
489         uint8_t hypercall[16];
490 
491         /* indicate KVM hypercall interface */
492         _FDT((fdt_begin_node(fdt, "hypervisor")));
493         _FDT((fdt_property_string(fdt, "compatible", "linux,kvm")));
494         if (kvmppc_has_cap_fixup_hcalls()) {
495             /*
496              * Older KVM versions with older guest kernels were broken with the
497              * magic page, don't allow the guest to map it.
498              */
499             kvmppc_get_hypercall(first_cpu->env_ptr, hypercall,
500                                  sizeof(hypercall));
501             _FDT((fdt_property(fdt, "hcall-instructions", hypercall,
502                               sizeof(hypercall))));
503         }
504         _FDT((fdt_end_node(fdt)));
505     }
506 
507     _FDT((fdt_end_node(fdt))); /* close root node */
508     _FDT((fdt_finish(fdt)));
509 
510     return fdt;
511 }
512 
513 static int spapr_populate_memory_node(void *fdt, int nodeid, hwaddr start,
514                                        hwaddr size)
515 {
516     uint32_t associativity[] = {
517         cpu_to_be32(0x4), /* length */
518         cpu_to_be32(0x0), cpu_to_be32(0x0),
519         cpu_to_be32(0x0), cpu_to_be32(nodeid)
520     };
521     char mem_name[32];
522     uint64_t mem_reg_property[2];
523     int off;
524 
525     mem_reg_property[0] = cpu_to_be64(start);
526     mem_reg_property[1] = cpu_to_be64(size);
527 
528     sprintf(mem_name, "memory@" TARGET_FMT_lx, start);
529     off = fdt_add_subnode(fdt, 0, mem_name);
530     _FDT(off);
531     _FDT((fdt_setprop_string(fdt, off, "device_type", "memory")));
532     _FDT((fdt_setprop(fdt, off, "reg", mem_reg_property,
533                       sizeof(mem_reg_property))));
534     _FDT((fdt_setprop(fdt, off, "ibm,associativity", associativity,
535                       sizeof(associativity))));
536     return off;
537 }
538 
539 static int spapr_populate_memory(sPAPRMachineState *spapr, void *fdt)
540 {
541     MachineState *machine = MACHINE(spapr);
542     hwaddr mem_start, node_size;
543     int i, nb_nodes = nb_numa_nodes;
544     NodeInfo *nodes = numa_info;
545     NodeInfo ramnode;
546 
547     /* No NUMA nodes, assume there is just one node with whole RAM */
548     if (!nb_numa_nodes) {
549         nb_nodes = 1;
550         ramnode.node_mem = machine->ram_size;
551         nodes = &ramnode;
552     }
553 
554     for (i = 0, mem_start = 0; i < nb_nodes; ++i) {
555         if (!nodes[i].node_mem) {
556             continue;
557         }
558         if (mem_start >= machine->ram_size) {
559             node_size = 0;
560         } else {
561             node_size = nodes[i].node_mem;
562             if (node_size > machine->ram_size - mem_start) {
563                 node_size = machine->ram_size - mem_start;
564             }
565         }
566         if (!mem_start) {
567             /* ppc_spapr_init() checks for rma_size <= node0_size already */
568             spapr_populate_memory_node(fdt, i, 0, spapr->rma_size);
569             mem_start += spapr->rma_size;
570             node_size -= spapr->rma_size;
571         }
572         for ( ; node_size; ) {
573             hwaddr sizetmp = pow2floor(node_size);
574 
575             /* mem_start != 0 here */
576             if (ctzl(mem_start) < ctzl(sizetmp)) {
577                 sizetmp = 1ULL << ctzl(mem_start);
578             }
579 
580             spapr_populate_memory_node(fdt, i, mem_start, sizetmp);
581             node_size -= sizetmp;
582             mem_start += sizetmp;
583         }
584     }
585 
586     return 0;
587 }
588 
589 static void spapr_populate_cpu_dt(CPUState *cs, void *fdt, int offset,
590                                   sPAPRMachineState *spapr)
591 {
592     PowerPCCPU *cpu = POWERPC_CPU(cs);
593     CPUPPCState *env = &cpu->env;
594     PowerPCCPUClass *pcc = POWERPC_CPU_GET_CLASS(cs);
595     int index = ppc_get_vcpu_dt_id(cpu);
596     uint32_t segs[] = {cpu_to_be32(28), cpu_to_be32(40),
597                        0xffffffff, 0xffffffff};
598     uint32_t tbfreq = kvm_enabled() ? kvmppc_get_tbfreq() : TIMEBASE_FREQ;
599     uint32_t cpufreq = kvm_enabled() ? kvmppc_get_clockfreq() : 1000000000;
600     uint32_t page_sizes_prop[64];
601     size_t page_sizes_prop_size;
602     uint32_t vcpus_per_socket = smp_threads * smp_cores;
603     uint32_t pft_size_prop[] = {0, cpu_to_be32(spapr->htab_shift)};
604 
605     /* Note: we keep CI large pages off for now because a 64K capable guest
606      * provisioned with large pages might otherwise try to map a qemu
607      * framebuffer (or other kind of memory mapped PCI BAR) using 64K pages
608      * even if that qemu runs on a 4k host.
609      *
610      * We can later add this bit back when we are confident this is not
611      * an issue (!HV KVM or 64K host)
612      */
613     uint8_t pa_features_206[] = { 6, 0,
614         0xf6, 0x1f, 0xc7, 0x00, 0x80, 0xc0 };
615     uint8_t pa_features_207[] = { 24, 0,
616         0xf6, 0x1f, 0xc7, 0xc0, 0x80, 0xf0,
617         0x80, 0x00, 0x00, 0x00, 0x00, 0x00,
618         0x00, 0x00, 0x00, 0x00, 0x80, 0x00,
619         0x80, 0x00, 0x80, 0x00, 0x80, 0x00 };
620     uint8_t *pa_features;
621     size_t pa_size;
622 
623     _FDT((fdt_setprop_cell(fdt, offset, "reg", index)));
624     _FDT((fdt_setprop_string(fdt, offset, "device_type", "cpu")));
625 
626     _FDT((fdt_setprop_cell(fdt, offset, "cpu-version", env->spr[SPR_PVR])));
627     _FDT((fdt_setprop_cell(fdt, offset, "d-cache-block-size",
628                            env->dcache_line_size)));
629     _FDT((fdt_setprop_cell(fdt, offset, "d-cache-line-size",
630                            env->dcache_line_size)));
631     _FDT((fdt_setprop_cell(fdt, offset, "i-cache-block-size",
632                            env->icache_line_size)));
633     _FDT((fdt_setprop_cell(fdt, offset, "i-cache-line-size",
634                            env->icache_line_size)));
635 
636     if (pcc->l1_dcache_size) {
637         _FDT((fdt_setprop_cell(fdt, offset, "d-cache-size",
638                                pcc->l1_dcache_size)));
639     } else {
640         fprintf(stderr, "Warning: Unknown L1 dcache size for cpu\n");
641     }
642     if (pcc->l1_icache_size) {
643         _FDT((fdt_setprop_cell(fdt, offset, "i-cache-size",
644                                pcc->l1_icache_size)));
645     } else {
646         fprintf(stderr, "Warning: Unknown L1 icache size for cpu\n");
647     }
648 
649     _FDT((fdt_setprop_cell(fdt, offset, "timebase-frequency", tbfreq)));
650     _FDT((fdt_setprop_cell(fdt, offset, "clock-frequency", cpufreq)));
651     _FDT((fdt_setprop_cell(fdt, offset, "slb-size", env->slb_nr)));
652     _FDT((fdt_setprop_cell(fdt, offset, "ibm,slb-size", env->slb_nr)));
653     _FDT((fdt_setprop_string(fdt, offset, "status", "okay")));
654     _FDT((fdt_setprop(fdt, offset, "64-bit", NULL, 0)));
655 
656     if (env->spr_cb[SPR_PURR].oea_read) {
657         _FDT((fdt_setprop(fdt, offset, "ibm,purr", NULL, 0)));
658     }
659 
660     if (env->mmu_model & POWERPC_MMU_1TSEG) {
661         _FDT((fdt_setprop(fdt, offset, "ibm,processor-segment-sizes",
662                           segs, sizeof(segs))));
663     }
664 
665     /* Advertise VMX/VSX (vector extensions) if available
666      *   0 / no property == no vector extensions
667      *   1               == VMX / Altivec available
668      *   2               == VSX available */
669     if (env->insns_flags & PPC_ALTIVEC) {
670         uint32_t vmx = (env->insns_flags2 & PPC2_VSX) ? 2 : 1;
671 
672         _FDT((fdt_setprop_cell(fdt, offset, "ibm,vmx", vmx)));
673     }
674 
675     /* Advertise DFP (Decimal Floating Point) if available
676      *   0 / no property == no DFP
677      *   1               == DFP available */
678     if (env->insns_flags2 & PPC2_DFP) {
679         _FDT((fdt_setprop_cell(fdt, offset, "ibm,dfp", 1)));
680     }
681 
682     page_sizes_prop_size = create_page_sizes_prop(env, page_sizes_prop,
683                                                   sizeof(page_sizes_prop));
684     if (page_sizes_prop_size) {
685         _FDT((fdt_setprop(fdt, offset, "ibm,segment-page-sizes",
686                           page_sizes_prop, page_sizes_prop_size)));
687     }
688 
689     /* Do the ibm,pa-features property, adjust it for ci-large-pages */
690     if (env->mmu_model == POWERPC_MMU_2_06) {
691         pa_features = pa_features_206;
692         pa_size = sizeof(pa_features_206);
693     } else /* env->mmu_model == POWERPC_MMU_2_07 */ {
694         pa_features = pa_features_207;
695         pa_size = sizeof(pa_features_207);
696     }
697     if (env->ci_large_pages) {
698         pa_features[3] |= 0x20;
699     }
700     _FDT((fdt_setprop(fdt, offset, "ibm,pa-features", pa_features, pa_size)));
701 
702     _FDT((fdt_setprop_cell(fdt, offset, "ibm,chip-id",
703                            cs->cpu_index / vcpus_per_socket)));
704 
705     _FDT((fdt_setprop(fdt, offset, "ibm,pft-size",
706                       pft_size_prop, sizeof(pft_size_prop))));
707 
708     _FDT(spapr_fixup_cpu_numa_dt(fdt, offset, cs));
709 
710     _FDT(spapr_fixup_cpu_smt_dt(fdt, offset, cpu,
711                                 ppc_get_compat_smt_threads(cpu)));
712 }
713 
714 static void spapr_populate_cpus_dt_node(void *fdt, sPAPRMachineState *spapr)
715 {
716     CPUState *cs;
717     int cpus_offset;
718     char *nodename;
719     int smt = kvmppc_smt_threads();
720 
721     cpus_offset = fdt_add_subnode(fdt, 0, "cpus");
722     _FDT(cpus_offset);
723     _FDT((fdt_setprop_cell(fdt, cpus_offset, "#address-cells", 0x1)));
724     _FDT((fdt_setprop_cell(fdt, cpus_offset, "#size-cells", 0x0)));
725 
726     /*
727      * We walk the CPUs in reverse order to ensure that CPU DT nodes
728      * created by fdt_add_subnode() end up in the right order in FDT
729      * for the guest kernel the enumerate the CPUs correctly.
730      */
731     CPU_FOREACH_REVERSE(cs) {
732         PowerPCCPU *cpu = POWERPC_CPU(cs);
733         int index = ppc_get_vcpu_dt_id(cpu);
734         DeviceClass *dc = DEVICE_GET_CLASS(cs);
735         int offset;
736 
737         if ((index % smt) != 0) {
738             continue;
739         }
740 
741         nodename = g_strdup_printf("%s@%x", dc->fw_name, index);
742         offset = fdt_add_subnode(fdt, cpus_offset, nodename);
743         g_free(nodename);
744         _FDT(offset);
745         spapr_populate_cpu_dt(cs, fdt, offset, spapr);
746     }
747 
748 }
749 
750 /*
751  * Adds ibm,dynamic-reconfiguration-memory node.
752  * Refer to docs/specs/ppc-spapr-hotplug.txt for the documentation
753  * of this device tree node.
754  */
755 static int spapr_populate_drconf_memory(sPAPRMachineState *spapr, void *fdt)
756 {
757     MachineState *machine = MACHINE(spapr);
758     int ret, i, offset;
759     uint64_t lmb_size = SPAPR_MEMORY_BLOCK_SIZE;
760     uint32_t prop_lmb_size[] = {0, cpu_to_be32(lmb_size)};
761     uint32_t nr_lmbs = (machine->maxram_size - machine->ram_size)/lmb_size;
762     uint32_t *int_buf, *cur_index, buf_len;
763     int nr_nodes = nb_numa_nodes ? nb_numa_nodes : 1;
764 
765     /*
766      * Allocate enough buffer size to fit in ibm,dynamic-memory
767      * or ibm,associativity-lookup-arrays
768      */
769     buf_len = MAX(nr_lmbs * SPAPR_DR_LMB_LIST_ENTRY_SIZE + 1, nr_nodes * 4 + 2)
770               * sizeof(uint32_t);
771     cur_index = int_buf = g_malloc0(buf_len);
772 
773     offset = fdt_add_subnode(fdt, 0, "ibm,dynamic-reconfiguration-memory");
774 
775     ret = fdt_setprop(fdt, offset, "ibm,lmb-size", prop_lmb_size,
776                     sizeof(prop_lmb_size));
777     if (ret < 0) {
778         goto out;
779     }
780 
781     ret = fdt_setprop_cell(fdt, offset, "ibm,memory-flags-mask", 0xff);
782     if (ret < 0) {
783         goto out;
784     }
785 
786     ret = fdt_setprop_cell(fdt, offset, "ibm,memory-preservation-time", 0x0);
787     if (ret < 0) {
788         goto out;
789     }
790 
791     /* ibm,dynamic-memory */
792     int_buf[0] = cpu_to_be32(nr_lmbs);
793     cur_index++;
794     for (i = 0; i < nr_lmbs; i++) {
795         sPAPRDRConnector *drc;
796         sPAPRDRConnectorClass *drck;
797         uint64_t addr = i * lmb_size + spapr->hotplug_memory.base;;
798         uint32_t *dynamic_memory = cur_index;
799 
800         drc = spapr_dr_connector_by_id(SPAPR_DR_CONNECTOR_TYPE_LMB,
801                                        addr/lmb_size);
802         g_assert(drc);
803         drck = SPAPR_DR_CONNECTOR_GET_CLASS(drc);
804 
805         dynamic_memory[0] = cpu_to_be32(addr >> 32);
806         dynamic_memory[1] = cpu_to_be32(addr & 0xffffffff);
807         dynamic_memory[2] = cpu_to_be32(drck->get_index(drc));
808         dynamic_memory[3] = cpu_to_be32(0); /* reserved */
809         dynamic_memory[4] = cpu_to_be32(numa_get_node(addr, NULL));
810         if (addr < machine->ram_size ||
811                     memory_region_present(get_system_memory(), addr)) {
812             dynamic_memory[5] = cpu_to_be32(SPAPR_LMB_FLAGS_ASSIGNED);
813         } else {
814             dynamic_memory[5] = cpu_to_be32(0);
815         }
816 
817         cur_index += SPAPR_DR_LMB_LIST_ENTRY_SIZE;
818     }
819     ret = fdt_setprop(fdt, offset, "ibm,dynamic-memory", int_buf, buf_len);
820     if (ret < 0) {
821         goto out;
822     }
823 
824     /* ibm,associativity-lookup-arrays */
825     cur_index = int_buf;
826     int_buf[0] = cpu_to_be32(nr_nodes);
827     int_buf[1] = cpu_to_be32(4); /* Number of entries per associativity list */
828     cur_index += 2;
829     for (i = 0; i < nr_nodes; i++) {
830         uint32_t associativity[] = {
831             cpu_to_be32(0x0),
832             cpu_to_be32(0x0),
833             cpu_to_be32(0x0),
834             cpu_to_be32(i)
835         };
836         memcpy(cur_index, associativity, sizeof(associativity));
837         cur_index += 4;
838     }
839     ret = fdt_setprop(fdt, offset, "ibm,associativity-lookup-arrays", int_buf,
840             (cur_index - int_buf) * sizeof(uint32_t));
841 out:
842     g_free(int_buf);
843     return ret;
844 }
845 
846 int spapr_h_cas_compose_response(sPAPRMachineState *spapr,
847                                  target_ulong addr, target_ulong size,
848                                  bool cpu_update, bool memory_update)
849 {
850     void *fdt, *fdt_skel;
851     sPAPRDeviceTreeUpdateHeader hdr = { .version_id = 1 };
852     sPAPRMachineClass *smc = SPAPR_MACHINE_GET_CLASS(qdev_get_machine());
853 
854     size -= sizeof(hdr);
855 
856     /* Create sceleton */
857     fdt_skel = g_malloc0(size);
858     _FDT((fdt_create(fdt_skel, size)));
859     _FDT((fdt_begin_node(fdt_skel, "")));
860     _FDT((fdt_end_node(fdt_skel)));
861     _FDT((fdt_finish(fdt_skel)));
862     fdt = g_malloc0(size);
863     _FDT((fdt_open_into(fdt_skel, fdt, size)));
864     g_free(fdt_skel);
865 
866     /* Fixup cpu nodes */
867     if (cpu_update) {
868         _FDT((spapr_fixup_cpu_dt(fdt, spapr)));
869     }
870 
871     /* Generate memory nodes or ibm,dynamic-reconfiguration-memory node */
872     if (memory_update && smc->dr_lmb_enabled) {
873         _FDT((spapr_populate_drconf_memory(spapr, fdt)));
874     }
875 
876     /* Pack resulting tree */
877     _FDT((fdt_pack(fdt)));
878 
879     if (fdt_totalsize(fdt) + sizeof(hdr) > size) {
880         trace_spapr_cas_failed(size);
881         return -1;
882     }
883 
884     cpu_physical_memory_write(addr, &hdr, sizeof(hdr));
885     cpu_physical_memory_write(addr + sizeof(hdr), fdt, fdt_totalsize(fdt));
886     trace_spapr_cas_continue(fdt_totalsize(fdt) + sizeof(hdr));
887     g_free(fdt);
888 
889     return 0;
890 }
891 
892 static void spapr_finalize_fdt(sPAPRMachineState *spapr,
893                                hwaddr fdt_addr,
894                                hwaddr rtas_addr,
895                                hwaddr rtas_size)
896 {
897     MachineState *machine = MACHINE(qdev_get_machine());
898     sPAPRMachineClass *smc = SPAPR_MACHINE_GET_CLASS(machine);
899     const char *boot_device = machine->boot_order;
900     int ret, i;
901     size_t cb = 0;
902     char *bootlist;
903     void *fdt;
904     sPAPRPHBState *phb;
905 
906     fdt = g_malloc(FDT_MAX_SIZE);
907 
908     /* open out the base tree into a temp buffer for the final tweaks */
909     _FDT((fdt_open_into(spapr->fdt_skel, fdt, FDT_MAX_SIZE)));
910 
911     ret = spapr_populate_memory(spapr, fdt);
912     if (ret < 0) {
913         fprintf(stderr, "couldn't setup memory nodes in fdt\n");
914         exit(1);
915     }
916 
917     ret = spapr_populate_vdevice(spapr->vio_bus, fdt);
918     if (ret < 0) {
919         fprintf(stderr, "couldn't setup vio devices in fdt\n");
920         exit(1);
921     }
922 
923     if (object_resolve_path_type("", TYPE_SPAPR_RNG, NULL)) {
924         ret = spapr_rng_populate_dt(fdt);
925         if (ret < 0) {
926             fprintf(stderr, "could not set up rng device in the fdt\n");
927             exit(1);
928         }
929     }
930 
931     QLIST_FOREACH(phb, &spapr->phbs, list) {
932         ret = spapr_populate_pci_dt(phb, PHANDLE_XICP, fdt);
933     }
934 
935     if (ret < 0) {
936         fprintf(stderr, "couldn't setup PCI devices in fdt\n");
937         exit(1);
938     }
939 
940     /* RTAS */
941     ret = spapr_rtas_device_tree_setup(fdt, rtas_addr, rtas_size);
942     if (ret < 0) {
943         fprintf(stderr, "Couldn't set up RTAS device tree properties\n");
944     }
945 
946     /* cpus */
947     spapr_populate_cpus_dt_node(fdt, spapr);
948 
949     bootlist = get_boot_devices_list(&cb, true);
950     if (cb && bootlist) {
951         int offset = fdt_path_offset(fdt, "/chosen");
952         if (offset < 0) {
953             exit(1);
954         }
955         for (i = 0; i < cb; i++) {
956             if (bootlist[i] == '\n') {
957                 bootlist[i] = ' ';
958             }
959 
960         }
961         ret = fdt_setprop_string(fdt, offset, "qemu,boot-list", bootlist);
962     }
963 
964     if (boot_device && strlen(boot_device)) {
965         int offset = fdt_path_offset(fdt, "/chosen");
966 
967         if (offset < 0) {
968             exit(1);
969         }
970         fdt_setprop_string(fdt, offset, "qemu,boot-device", boot_device);
971     }
972 
973     if (!spapr->has_graphics) {
974         spapr_populate_chosen_stdout(fdt, spapr->vio_bus);
975     }
976 
977     if (smc->dr_lmb_enabled) {
978         _FDT(spapr_drc_populate_dt(fdt, 0, NULL, SPAPR_DR_CONNECTOR_TYPE_LMB));
979     }
980 
981     _FDT((fdt_pack(fdt)));
982 
983     if (fdt_totalsize(fdt) > FDT_MAX_SIZE) {
984         error_report("FDT too big ! 0x%x bytes (max is 0x%x)",
985                      fdt_totalsize(fdt), FDT_MAX_SIZE);
986         exit(1);
987     }
988 
989     qemu_fdt_dumpdtb(fdt, fdt_totalsize(fdt));
990     cpu_physical_memory_write(fdt_addr, fdt, fdt_totalsize(fdt));
991 
992     g_free(bootlist);
993     g_free(fdt);
994 }
995 
996 static uint64_t translate_kernel_address(void *opaque, uint64_t addr)
997 {
998     return (addr & 0x0fffffff) + KERNEL_LOAD_ADDR;
999 }
1000 
1001 static void emulate_spapr_hypercall(PowerPCCPU *cpu)
1002 {
1003     CPUPPCState *env = &cpu->env;
1004 
1005     if (msr_pr) {
1006         hcall_dprintf("Hypercall made with MSR[PR]=1\n");
1007         env->gpr[3] = H_PRIVILEGE;
1008     } else {
1009         env->gpr[3] = spapr_hypercall(cpu, env->gpr[3], &env->gpr[4]);
1010     }
1011 }
1012 
1013 #define HPTE(_table, _i)   (void *)(((uint64_t *)(_table)) + ((_i) * 2))
1014 #define HPTE_VALID(_hpte)  (tswap64(*((uint64_t *)(_hpte))) & HPTE64_V_VALID)
1015 #define HPTE_DIRTY(_hpte)  (tswap64(*((uint64_t *)(_hpte))) & HPTE64_V_HPTE_DIRTY)
1016 #define CLEAN_HPTE(_hpte)  ((*(uint64_t *)(_hpte)) &= tswap64(~HPTE64_V_HPTE_DIRTY))
1017 #define DIRTY_HPTE(_hpte)  ((*(uint64_t *)(_hpte)) |= tswap64(HPTE64_V_HPTE_DIRTY))
1018 
1019 static void spapr_alloc_htab(sPAPRMachineState *spapr)
1020 {
1021     long shift;
1022     int index;
1023 
1024     /* allocate hash page table.  For now we always make this 16mb,
1025      * later we should probably make it scale to the size of guest
1026      * RAM */
1027 
1028     shift = kvmppc_reset_htab(spapr->htab_shift);
1029     if (shift < 0) {
1030         /*
1031          * For HV KVM, host kernel will return -ENOMEM when requested
1032          * HTAB size can't be allocated.
1033          */
1034         error_setg(&error_abort, "Failed to allocate HTAB of requested size, try with smaller maxmem");
1035     } else if (shift > 0) {
1036         /*
1037          * Kernel handles htab, we don't need to allocate one
1038          *
1039          * Older kernels can fall back to lower HTAB shift values,
1040          * but we don't allow booting of such guests.
1041          */
1042         if (shift != spapr->htab_shift) {
1043             error_setg(&error_abort, "Failed to allocate HTAB of requested size, try with smaller maxmem");
1044         }
1045 
1046         spapr->htab_shift = shift;
1047         kvmppc_kern_htab = true;
1048     } else {
1049         /* Allocate htab */
1050         spapr->htab = qemu_memalign(HTAB_SIZE(spapr), HTAB_SIZE(spapr));
1051 
1052         /* And clear it */
1053         memset(spapr->htab, 0, HTAB_SIZE(spapr));
1054 
1055         for (index = 0; index < HTAB_SIZE(spapr) / HASH_PTE_SIZE_64; index++) {
1056             DIRTY_HPTE(HPTE(spapr->htab, index));
1057         }
1058     }
1059 }
1060 
1061 /*
1062  * Clear HTAB entries during reset.
1063  *
1064  * If host kernel has allocated HTAB, KVM_PPC_ALLOCATE_HTAB ioctl is
1065  * used to clear HTAB. Otherwise QEMU-allocated HTAB is cleared manually.
1066  */
1067 static void spapr_reset_htab(sPAPRMachineState *spapr)
1068 {
1069     long shift;
1070     int index;
1071 
1072     shift = kvmppc_reset_htab(spapr->htab_shift);
1073     if (shift < 0) {
1074         error_setg(&error_abort, "Failed to reset HTAB");
1075     } else if (shift > 0) {
1076         if (shift != spapr->htab_shift) {
1077             error_setg(&error_abort, "Requested HTAB allocation failed during reset");
1078         }
1079 
1080         /* Tell readers to update their file descriptor */
1081         if (spapr->htab_fd >= 0) {
1082             spapr->htab_fd_stale = true;
1083         }
1084     } else {
1085         memset(spapr->htab, 0, HTAB_SIZE(spapr));
1086 
1087         for (index = 0; index < HTAB_SIZE(spapr) / HASH_PTE_SIZE_64; index++) {
1088             DIRTY_HPTE(HPTE(spapr->htab, index));
1089         }
1090     }
1091 
1092     /* Update the RMA size if necessary */
1093     if (spapr->vrma_adjust) {
1094         spapr->rma_size = kvmppc_rma_size(spapr_node0_size(),
1095                                           spapr->htab_shift);
1096     }
1097 }
1098 
1099 static int find_unknown_sysbus_device(SysBusDevice *sbdev, void *opaque)
1100 {
1101     bool matched = false;
1102 
1103     if (object_dynamic_cast(OBJECT(sbdev), TYPE_SPAPR_PCI_HOST_BRIDGE)) {
1104         matched = true;
1105     }
1106 
1107     if (!matched) {
1108         error_report("Device %s is not supported by this machine yet.",
1109                      qdev_fw_name(DEVICE(sbdev)));
1110         exit(1);
1111     }
1112 
1113     return 0;
1114 }
1115 
1116 /*
1117  * A guest reset will cause spapr->htab_fd to become stale if being used.
1118  * Reopen the file descriptor to make sure the whole HTAB is properly read.
1119  */
1120 static int spapr_check_htab_fd(sPAPRMachineState *spapr)
1121 {
1122     int rc = 0;
1123 
1124     if (spapr->htab_fd_stale) {
1125         close(spapr->htab_fd);
1126         spapr->htab_fd = kvmppc_get_htab_fd(false);
1127         if (spapr->htab_fd < 0) {
1128             error_report("Unable to open fd for reading hash table from KVM: "
1129                          "%s", strerror(errno));
1130             rc = -1;
1131         }
1132         spapr->htab_fd_stale = false;
1133     }
1134 
1135     return rc;
1136 }
1137 
1138 static void ppc_spapr_reset(void)
1139 {
1140     sPAPRMachineState *spapr = SPAPR_MACHINE(qdev_get_machine());
1141     PowerPCCPU *first_ppc_cpu;
1142     uint32_t rtas_limit;
1143 
1144     /* Check for unknown sysbus devices */
1145     foreach_dynamic_sysbus_device(find_unknown_sysbus_device, NULL);
1146 
1147     /* Reset the hash table & recalc the RMA */
1148     spapr_reset_htab(spapr);
1149 
1150     qemu_devices_reset();
1151 
1152     /*
1153      * We place the device tree and RTAS just below either the top of the RMA,
1154      * or just below 2GB, whichever is lowere, so that it can be
1155      * processed with 32-bit real mode code if necessary
1156      */
1157     rtas_limit = MIN(spapr->rma_size, RTAS_MAX_ADDR);
1158     spapr->rtas_addr = rtas_limit - RTAS_MAX_SIZE;
1159     spapr->fdt_addr = spapr->rtas_addr - FDT_MAX_SIZE;
1160 
1161     /* Load the fdt */
1162     spapr_finalize_fdt(spapr, spapr->fdt_addr, spapr->rtas_addr,
1163                        spapr->rtas_size);
1164 
1165     /* Copy RTAS over */
1166     cpu_physical_memory_write(spapr->rtas_addr, spapr->rtas_blob,
1167                               spapr->rtas_size);
1168 
1169     /* Set up the entry state */
1170     first_ppc_cpu = POWERPC_CPU(first_cpu);
1171     first_ppc_cpu->env.gpr[3] = spapr->fdt_addr;
1172     first_ppc_cpu->env.gpr[5] = 0;
1173     first_cpu->halted = 0;
1174     first_ppc_cpu->env.nip = SPAPR_ENTRY_POINT;
1175 
1176 }
1177 
1178 static void spapr_cpu_reset(void *opaque)
1179 {
1180     sPAPRMachineState *spapr = SPAPR_MACHINE(qdev_get_machine());
1181     PowerPCCPU *cpu = opaque;
1182     CPUState *cs = CPU(cpu);
1183     CPUPPCState *env = &cpu->env;
1184 
1185     cpu_reset(cs);
1186 
1187     /* All CPUs start halted.  CPU0 is unhalted from the machine level
1188      * reset code and the rest are explicitly started up by the guest
1189      * using an RTAS call */
1190     cs->halted = 1;
1191 
1192     env->spr[SPR_HIOR] = 0;
1193 
1194     env->external_htab = (uint8_t *)spapr->htab;
1195     if (kvm_enabled() && !env->external_htab) {
1196         /*
1197          * HV KVM, set external_htab to 1 so our ppc_hash64_load_hpte*
1198          * functions do the right thing.
1199          */
1200         env->external_htab = (void *)1;
1201     }
1202     env->htab_base = -1;
1203     /*
1204      * htab_mask is the mask used to normalize hash value to PTEG index.
1205      * htab_shift is log2 of hash table size.
1206      * We have 8 hpte per group, and each hpte is 16 bytes.
1207      * ie have 128 bytes per hpte entry.
1208      */
1209     env->htab_mask = (1ULL << (spapr->htab_shift - 7)) - 1;
1210     env->spr[SPR_SDR1] = (target_ulong)(uintptr_t)spapr->htab |
1211         (spapr->htab_shift - 18);
1212 }
1213 
1214 static void spapr_create_nvram(sPAPRMachineState *spapr)
1215 {
1216     DeviceState *dev = qdev_create(&spapr->vio_bus->bus, "spapr-nvram");
1217     DriveInfo *dinfo = drive_get(IF_PFLASH, 0, 0);
1218 
1219     if (dinfo) {
1220         qdev_prop_set_drive(dev, "drive", blk_by_legacy_dinfo(dinfo),
1221                             &error_fatal);
1222     }
1223 
1224     qdev_init_nofail(dev);
1225 
1226     spapr->nvram = (struct sPAPRNVRAM *)dev;
1227 }
1228 
1229 static void spapr_rtc_create(sPAPRMachineState *spapr)
1230 {
1231     DeviceState *dev = qdev_create(NULL, TYPE_SPAPR_RTC);
1232 
1233     qdev_init_nofail(dev);
1234     spapr->rtc = dev;
1235 
1236     object_property_add_alias(qdev_get_machine(), "rtc-time",
1237                               OBJECT(spapr->rtc), "date", NULL);
1238 }
1239 
1240 /* Returns whether we want to use VGA or not */
1241 static int spapr_vga_init(PCIBus *pci_bus)
1242 {
1243     switch (vga_interface_type) {
1244     case VGA_NONE:
1245         return false;
1246     case VGA_DEVICE:
1247         return true;
1248     case VGA_STD:
1249     case VGA_VIRTIO:
1250         return pci_vga_init(pci_bus) != NULL;
1251     default:
1252         fprintf(stderr, "This vga model is not supported,"
1253                 "currently it only supports -vga std\n");
1254         exit(0);
1255     }
1256 }
1257 
1258 static int spapr_post_load(void *opaque, int version_id)
1259 {
1260     sPAPRMachineState *spapr = (sPAPRMachineState *)opaque;
1261     int err = 0;
1262 
1263     /* In earlier versions, there was no separate qdev for the PAPR
1264      * RTC, so the RTC offset was stored directly in sPAPREnvironment.
1265      * So when migrating from those versions, poke the incoming offset
1266      * value into the RTC device */
1267     if (version_id < 3) {
1268         err = spapr_rtc_import_offset(spapr->rtc, spapr->rtc_offset);
1269     }
1270 
1271     return err;
1272 }
1273 
1274 static bool version_before_3(void *opaque, int version_id)
1275 {
1276     return version_id < 3;
1277 }
1278 
1279 static const VMStateDescription vmstate_spapr = {
1280     .name = "spapr",
1281     .version_id = 3,
1282     .minimum_version_id = 1,
1283     .post_load = spapr_post_load,
1284     .fields = (VMStateField[]) {
1285         /* used to be @next_irq */
1286         VMSTATE_UNUSED_BUFFER(version_before_3, 0, 4),
1287 
1288         /* RTC offset */
1289         VMSTATE_UINT64_TEST(rtc_offset, sPAPRMachineState, version_before_3),
1290 
1291         VMSTATE_PPC_TIMEBASE_V(tb, sPAPRMachineState, 2),
1292         VMSTATE_END_OF_LIST()
1293     },
1294 };
1295 
1296 static int htab_save_setup(QEMUFile *f, void *opaque)
1297 {
1298     sPAPRMachineState *spapr = opaque;
1299 
1300     /* "Iteration" header */
1301     qemu_put_be32(f, spapr->htab_shift);
1302 
1303     if (spapr->htab) {
1304         spapr->htab_save_index = 0;
1305         spapr->htab_first_pass = true;
1306     } else {
1307         assert(kvm_enabled());
1308 
1309         spapr->htab_fd = kvmppc_get_htab_fd(false);
1310         spapr->htab_fd_stale = false;
1311         if (spapr->htab_fd < 0) {
1312             fprintf(stderr, "Unable to open fd for reading hash table from KVM: %s\n",
1313                     strerror(errno));
1314             return -1;
1315         }
1316     }
1317 
1318 
1319     return 0;
1320 }
1321 
1322 static void htab_save_first_pass(QEMUFile *f, sPAPRMachineState *spapr,
1323                                  int64_t max_ns)
1324 {
1325     int htabslots = HTAB_SIZE(spapr) / HASH_PTE_SIZE_64;
1326     int index = spapr->htab_save_index;
1327     int64_t starttime = qemu_clock_get_ns(QEMU_CLOCK_REALTIME);
1328 
1329     assert(spapr->htab_first_pass);
1330 
1331     do {
1332         int chunkstart;
1333 
1334         /* Consume invalid HPTEs */
1335         while ((index < htabslots)
1336                && !HPTE_VALID(HPTE(spapr->htab, index))) {
1337             index++;
1338             CLEAN_HPTE(HPTE(spapr->htab, index));
1339         }
1340 
1341         /* Consume valid HPTEs */
1342         chunkstart = index;
1343         while ((index < htabslots) && (index - chunkstart < USHRT_MAX)
1344                && HPTE_VALID(HPTE(spapr->htab, index))) {
1345             index++;
1346             CLEAN_HPTE(HPTE(spapr->htab, index));
1347         }
1348 
1349         if (index > chunkstart) {
1350             int n_valid = index - chunkstart;
1351 
1352             qemu_put_be32(f, chunkstart);
1353             qemu_put_be16(f, n_valid);
1354             qemu_put_be16(f, 0);
1355             qemu_put_buffer(f, HPTE(spapr->htab, chunkstart),
1356                             HASH_PTE_SIZE_64 * n_valid);
1357 
1358             if ((qemu_clock_get_ns(QEMU_CLOCK_REALTIME) - starttime) > max_ns) {
1359                 break;
1360             }
1361         }
1362     } while ((index < htabslots) && !qemu_file_rate_limit(f));
1363 
1364     if (index >= htabslots) {
1365         assert(index == htabslots);
1366         index = 0;
1367         spapr->htab_first_pass = false;
1368     }
1369     spapr->htab_save_index = index;
1370 }
1371 
1372 static int htab_save_later_pass(QEMUFile *f, sPAPRMachineState *spapr,
1373                                 int64_t max_ns)
1374 {
1375     bool final = max_ns < 0;
1376     int htabslots = HTAB_SIZE(spapr) / HASH_PTE_SIZE_64;
1377     int examined = 0, sent = 0;
1378     int index = spapr->htab_save_index;
1379     int64_t starttime = qemu_clock_get_ns(QEMU_CLOCK_REALTIME);
1380 
1381     assert(!spapr->htab_first_pass);
1382 
1383     do {
1384         int chunkstart, invalidstart;
1385 
1386         /* Consume non-dirty HPTEs */
1387         while ((index < htabslots)
1388                && !HPTE_DIRTY(HPTE(spapr->htab, index))) {
1389             index++;
1390             examined++;
1391         }
1392 
1393         chunkstart = index;
1394         /* Consume valid dirty HPTEs */
1395         while ((index < htabslots) && (index - chunkstart < USHRT_MAX)
1396                && HPTE_DIRTY(HPTE(spapr->htab, index))
1397                && HPTE_VALID(HPTE(spapr->htab, index))) {
1398             CLEAN_HPTE(HPTE(spapr->htab, index));
1399             index++;
1400             examined++;
1401         }
1402 
1403         invalidstart = index;
1404         /* Consume invalid dirty HPTEs */
1405         while ((index < htabslots) && (index - invalidstart < USHRT_MAX)
1406                && HPTE_DIRTY(HPTE(spapr->htab, index))
1407                && !HPTE_VALID(HPTE(spapr->htab, index))) {
1408             CLEAN_HPTE(HPTE(spapr->htab, index));
1409             index++;
1410             examined++;
1411         }
1412 
1413         if (index > chunkstart) {
1414             int n_valid = invalidstart - chunkstart;
1415             int n_invalid = index - invalidstart;
1416 
1417             qemu_put_be32(f, chunkstart);
1418             qemu_put_be16(f, n_valid);
1419             qemu_put_be16(f, n_invalid);
1420             qemu_put_buffer(f, HPTE(spapr->htab, chunkstart),
1421                             HASH_PTE_SIZE_64 * n_valid);
1422             sent += index - chunkstart;
1423 
1424             if (!final && (qemu_clock_get_ns(QEMU_CLOCK_REALTIME) - starttime) > max_ns) {
1425                 break;
1426             }
1427         }
1428 
1429         if (examined >= htabslots) {
1430             break;
1431         }
1432 
1433         if (index >= htabslots) {
1434             assert(index == htabslots);
1435             index = 0;
1436         }
1437     } while ((examined < htabslots) && (!qemu_file_rate_limit(f) || final));
1438 
1439     if (index >= htabslots) {
1440         assert(index == htabslots);
1441         index = 0;
1442     }
1443 
1444     spapr->htab_save_index = index;
1445 
1446     return (examined >= htabslots) && (sent == 0) ? 1 : 0;
1447 }
1448 
1449 #define MAX_ITERATION_NS    5000000 /* 5 ms */
1450 #define MAX_KVM_BUF_SIZE    2048
1451 
1452 static int htab_save_iterate(QEMUFile *f, void *opaque)
1453 {
1454     sPAPRMachineState *spapr = opaque;
1455     int rc = 0;
1456 
1457     /* Iteration header */
1458     qemu_put_be32(f, 0);
1459 
1460     if (!spapr->htab) {
1461         assert(kvm_enabled());
1462 
1463         rc = spapr_check_htab_fd(spapr);
1464         if (rc < 0) {
1465             return rc;
1466         }
1467 
1468         rc = kvmppc_save_htab(f, spapr->htab_fd,
1469                               MAX_KVM_BUF_SIZE, MAX_ITERATION_NS);
1470         if (rc < 0) {
1471             return rc;
1472         }
1473     } else  if (spapr->htab_first_pass) {
1474         htab_save_first_pass(f, spapr, MAX_ITERATION_NS);
1475     } else {
1476         rc = htab_save_later_pass(f, spapr, MAX_ITERATION_NS);
1477     }
1478 
1479     /* End marker */
1480     qemu_put_be32(f, 0);
1481     qemu_put_be16(f, 0);
1482     qemu_put_be16(f, 0);
1483 
1484     return rc;
1485 }
1486 
1487 static int htab_save_complete(QEMUFile *f, void *opaque)
1488 {
1489     sPAPRMachineState *spapr = opaque;
1490 
1491     /* Iteration header */
1492     qemu_put_be32(f, 0);
1493 
1494     if (!spapr->htab) {
1495         int rc;
1496 
1497         assert(kvm_enabled());
1498 
1499         rc = spapr_check_htab_fd(spapr);
1500         if (rc < 0) {
1501             return rc;
1502         }
1503 
1504         rc = kvmppc_save_htab(f, spapr->htab_fd, MAX_KVM_BUF_SIZE, -1);
1505         if (rc < 0) {
1506             return rc;
1507         }
1508         close(spapr->htab_fd);
1509         spapr->htab_fd = -1;
1510     } else {
1511         htab_save_later_pass(f, spapr, -1);
1512     }
1513 
1514     /* End marker */
1515     qemu_put_be32(f, 0);
1516     qemu_put_be16(f, 0);
1517     qemu_put_be16(f, 0);
1518 
1519     return 0;
1520 }
1521 
1522 static int htab_load(QEMUFile *f, void *opaque, int version_id)
1523 {
1524     sPAPRMachineState *spapr = opaque;
1525     uint32_t section_hdr;
1526     int fd = -1;
1527 
1528     if (version_id < 1 || version_id > 1) {
1529         fprintf(stderr, "htab_load() bad version\n");
1530         return -EINVAL;
1531     }
1532 
1533     section_hdr = qemu_get_be32(f);
1534 
1535     if (section_hdr) {
1536         /* First section, just the hash shift */
1537         if (spapr->htab_shift != section_hdr) {
1538             error_report("htab_shift mismatch: source %d target %d",
1539                          section_hdr, spapr->htab_shift);
1540             return -EINVAL;
1541         }
1542         return 0;
1543     }
1544 
1545     if (!spapr->htab) {
1546         assert(kvm_enabled());
1547 
1548         fd = kvmppc_get_htab_fd(true);
1549         if (fd < 0) {
1550             fprintf(stderr, "Unable to open fd to restore KVM hash table: %s\n",
1551                     strerror(errno));
1552         }
1553     }
1554 
1555     while (true) {
1556         uint32_t index;
1557         uint16_t n_valid, n_invalid;
1558 
1559         index = qemu_get_be32(f);
1560         n_valid = qemu_get_be16(f);
1561         n_invalid = qemu_get_be16(f);
1562 
1563         if ((index == 0) && (n_valid == 0) && (n_invalid == 0)) {
1564             /* End of Stream */
1565             break;
1566         }
1567 
1568         if ((index + n_valid + n_invalid) >
1569             (HTAB_SIZE(spapr) / HASH_PTE_SIZE_64)) {
1570             /* Bad index in stream */
1571             fprintf(stderr, "htab_load() bad index %d (%hd+%hd entries) "
1572                     "in htab stream (htab_shift=%d)\n", index, n_valid, n_invalid,
1573                     spapr->htab_shift);
1574             return -EINVAL;
1575         }
1576 
1577         if (spapr->htab) {
1578             if (n_valid) {
1579                 qemu_get_buffer(f, HPTE(spapr->htab, index),
1580                                 HASH_PTE_SIZE_64 * n_valid);
1581             }
1582             if (n_invalid) {
1583                 memset(HPTE(spapr->htab, index + n_valid), 0,
1584                        HASH_PTE_SIZE_64 * n_invalid);
1585             }
1586         } else {
1587             int rc;
1588 
1589             assert(fd >= 0);
1590 
1591             rc = kvmppc_load_htab_chunk(f, fd, index, n_valid, n_invalid);
1592             if (rc < 0) {
1593                 return rc;
1594             }
1595         }
1596     }
1597 
1598     if (!spapr->htab) {
1599         assert(fd >= 0);
1600         close(fd);
1601     }
1602 
1603     return 0;
1604 }
1605 
1606 static SaveVMHandlers savevm_htab_handlers = {
1607     .save_live_setup = htab_save_setup,
1608     .save_live_iterate = htab_save_iterate,
1609     .save_live_complete_precopy = htab_save_complete,
1610     .load_state = htab_load,
1611 };
1612 
1613 static void spapr_boot_set(void *opaque, const char *boot_device,
1614                            Error **errp)
1615 {
1616     MachineState *machine = MACHINE(qdev_get_machine());
1617     machine->boot_order = g_strdup(boot_device);
1618 }
1619 
1620 static void spapr_cpu_init(sPAPRMachineState *spapr, PowerPCCPU *cpu)
1621 {
1622     CPUPPCState *env = &cpu->env;
1623 
1624     /* Set time-base frequency to 512 MHz */
1625     cpu_ppc_tb_init(env, TIMEBASE_FREQ);
1626 
1627     /* PAPR always has exception vectors in RAM not ROM. To ensure this,
1628      * MSR[IP] should never be set.
1629      */
1630     env->msr_mask &= ~(1 << 6);
1631 
1632     /* Tell KVM that we're in PAPR mode */
1633     if (kvm_enabled()) {
1634         kvmppc_set_papr(cpu);
1635     }
1636 
1637     if (cpu->max_compat) {
1638         if (ppc_set_compat(cpu, cpu->max_compat) < 0) {
1639             exit(1);
1640         }
1641     }
1642 
1643     xics_cpu_setup(spapr->icp, cpu);
1644 
1645     qemu_register_reset(spapr_cpu_reset, cpu);
1646 }
1647 
1648 /*
1649  * Reset routine for LMB DR devices.
1650  *
1651  * Unlike PCI DR devices, LMB DR devices explicitly register this reset
1652  * routine. Reset for PCI DR devices will be handled by PHB reset routine
1653  * when it walks all its children devices. LMB devices reset occurs
1654  * as part of spapr_ppc_reset().
1655  */
1656 static void spapr_drc_reset(void *opaque)
1657 {
1658     sPAPRDRConnector *drc = opaque;
1659     DeviceState *d = DEVICE(drc);
1660 
1661     if (d) {
1662         device_reset(d);
1663     }
1664 }
1665 
1666 static void spapr_create_lmb_dr_connectors(sPAPRMachineState *spapr)
1667 {
1668     MachineState *machine = MACHINE(spapr);
1669     uint64_t lmb_size = SPAPR_MEMORY_BLOCK_SIZE;
1670     uint32_t nr_lmbs = (machine->maxram_size - machine->ram_size)/lmb_size;
1671     int i;
1672 
1673     for (i = 0; i < nr_lmbs; i++) {
1674         sPAPRDRConnector *drc;
1675         uint64_t addr;
1676 
1677         addr = i * lmb_size + spapr->hotplug_memory.base;
1678         drc = spapr_dr_connector_new(OBJECT(spapr), SPAPR_DR_CONNECTOR_TYPE_LMB,
1679                                      addr/lmb_size);
1680         qemu_register_reset(spapr_drc_reset, drc);
1681     }
1682 }
1683 
1684 /*
1685  * If RAM size, maxmem size and individual node mem sizes aren't aligned
1686  * to SPAPR_MEMORY_BLOCK_SIZE(256MB), then refuse to start the guest
1687  * since we can't support such unaligned sizes with DRCONF_MEMORY.
1688  */
1689 static void spapr_validate_node_memory(MachineState *machine)
1690 {
1691     int i;
1692 
1693     if (machine->maxram_size % SPAPR_MEMORY_BLOCK_SIZE ||
1694         machine->ram_size % SPAPR_MEMORY_BLOCK_SIZE) {
1695         error_report("Can't support memory configuration where RAM size "
1696                      "0x" RAM_ADDR_FMT " or maxmem size "
1697                      "0x" RAM_ADDR_FMT " isn't aligned to %llu MB",
1698                      machine->ram_size, machine->maxram_size,
1699                      SPAPR_MEMORY_BLOCK_SIZE/M_BYTE);
1700         exit(EXIT_FAILURE);
1701     }
1702 
1703     for (i = 0; i < nb_numa_nodes; i++) {
1704         if (numa_info[i].node_mem % SPAPR_MEMORY_BLOCK_SIZE) {
1705             error_report("Can't support memory configuration where memory size"
1706                          " %" PRIx64 " of node %d isn't aligned to %llu MB",
1707                          numa_info[i].node_mem, i,
1708                          SPAPR_MEMORY_BLOCK_SIZE/M_BYTE);
1709             exit(EXIT_FAILURE);
1710         }
1711     }
1712 }
1713 
1714 /* pSeries LPAR / sPAPR hardware init */
1715 static void ppc_spapr_init(MachineState *machine)
1716 {
1717     sPAPRMachineState *spapr = SPAPR_MACHINE(machine);
1718     sPAPRMachineClass *smc = SPAPR_MACHINE_GET_CLASS(machine);
1719     const char *kernel_filename = machine->kernel_filename;
1720     const char *kernel_cmdline = machine->kernel_cmdline;
1721     const char *initrd_filename = machine->initrd_filename;
1722     PowerPCCPU *cpu;
1723     PCIHostState *phb;
1724     int i;
1725     MemoryRegion *sysmem = get_system_memory();
1726     MemoryRegion *ram = g_new(MemoryRegion, 1);
1727     MemoryRegion *rma_region;
1728     void *rma = NULL;
1729     hwaddr rma_alloc_size;
1730     hwaddr node0_size = spapr_node0_size();
1731     uint32_t initrd_base = 0;
1732     long kernel_size = 0, initrd_size = 0;
1733     long load_limit, fw_size;
1734     bool kernel_le = false;
1735     char *filename;
1736 
1737     msi_supported = true;
1738 
1739     QLIST_INIT(&spapr->phbs);
1740 
1741     cpu_ppc_hypercall = emulate_spapr_hypercall;
1742 
1743     /* Allocate RMA if necessary */
1744     rma_alloc_size = kvmppc_alloc_rma(&rma);
1745 
1746     if (rma_alloc_size == -1) {
1747         error_report("Unable to create RMA");
1748         exit(1);
1749     }
1750 
1751     if (rma_alloc_size && (rma_alloc_size < node0_size)) {
1752         spapr->rma_size = rma_alloc_size;
1753     } else {
1754         spapr->rma_size = node0_size;
1755 
1756         /* With KVM, we don't actually know whether KVM supports an
1757          * unbounded RMA (PR KVM) or is limited by the hash table size
1758          * (HV KVM using VRMA), so we always assume the latter
1759          *
1760          * In that case, we also limit the initial allocations for RTAS
1761          * etc... to 256M since we have no way to know what the VRMA size
1762          * is going to be as it depends on the size of the hash table
1763          * isn't determined yet.
1764          */
1765         if (kvm_enabled()) {
1766             spapr->vrma_adjust = 1;
1767             spapr->rma_size = MIN(spapr->rma_size, 0x10000000);
1768         }
1769     }
1770 
1771     if (spapr->rma_size > node0_size) {
1772         fprintf(stderr, "Error: Numa node 0 has to span the RMA (%#08"HWADDR_PRIx")\n",
1773                 spapr->rma_size);
1774         exit(1);
1775     }
1776 
1777     /* Setup a load limit for the ramdisk leaving room for SLOF and FDT */
1778     load_limit = MIN(spapr->rma_size, RTAS_MAX_ADDR) - FW_OVERHEAD;
1779 
1780     /* We aim for a hash table of size 1/128 the size of RAM.  The
1781      * normal rule of thumb is 1/64 the size of RAM, but that's much
1782      * more than needed for the Linux guests we support. */
1783     spapr->htab_shift = 18; /* Minimum architected size */
1784     while (spapr->htab_shift <= 46) {
1785         if ((1ULL << (spapr->htab_shift + 7)) >= machine->maxram_size) {
1786             break;
1787         }
1788         spapr->htab_shift++;
1789     }
1790     spapr_alloc_htab(spapr);
1791 
1792     /* Set up Interrupt Controller before we create the VCPUs */
1793     spapr->icp = xics_system_init(machine,
1794                                   DIV_ROUND_UP(max_cpus * kvmppc_smt_threads(),
1795                                                smp_threads),
1796                                   XICS_IRQS);
1797 
1798     if (smc->dr_lmb_enabled) {
1799         spapr_validate_node_memory(machine);
1800     }
1801 
1802     /* init CPUs */
1803     if (machine->cpu_model == NULL) {
1804         machine->cpu_model = kvm_enabled() ? "host" : "POWER7";
1805     }
1806     for (i = 0; i < smp_cpus; i++) {
1807         cpu = cpu_ppc_init(machine->cpu_model);
1808         if (cpu == NULL) {
1809             fprintf(stderr, "Unable to find PowerPC CPU definition\n");
1810             exit(1);
1811         }
1812         spapr_cpu_init(spapr, cpu);
1813     }
1814 
1815     if (kvm_enabled()) {
1816         /* Enable H_LOGICAL_CI_* so SLOF can talk to in-kernel devices */
1817         kvmppc_enable_logical_ci_hcalls();
1818         kvmppc_enable_set_mode_hcall();
1819     }
1820 
1821     /* allocate RAM */
1822     memory_region_allocate_system_memory(ram, NULL, "ppc_spapr.ram",
1823                                          machine->ram_size);
1824     memory_region_add_subregion(sysmem, 0, ram);
1825 
1826     if (rma_alloc_size && rma) {
1827         rma_region = g_new(MemoryRegion, 1);
1828         memory_region_init_ram_ptr(rma_region, NULL, "ppc_spapr.rma",
1829                                    rma_alloc_size, rma);
1830         vmstate_register_ram_global(rma_region);
1831         memory_region_add_subregion(sysmem, 0, rma_region);
1832     }
1833 
1834     /* initialize hotplug memory address space */
1835     if (machine->ram_size < machine->maxram_size) {
1836         ram_addr_t hotplug_mem_size = machine->maxram_size - machine->ram_size;
1837 
1838         if (machine->ram_slots > SPAPR_MAX_RAM_SLOTS) {
1839             error_report("Specified number of memory slots %" PRIu64
1840                          " exceeds max supported %d",
1841                          machine->ram_slots, SPAPR_MAX_RAM_SLOTS);
1842             exit(EXIT_FAILURE);
1843         }
1844 
1845         spapr->hotplug_memory.base = ROUND_UP(machine->ram_size,
1846                                               SPAPR_HOTPLUG_MEM_ALIGN);
1847         memory_region_init(&spapr->hotplug_memory.mr, OBJECT(spapr),
1848                            "hotplug-memory", hotplug_mem_size);
1849         memory_region_add_subregion(sysmem, spapr->hotplug_memory.base,
1850                                     &spapr->hotplug_memory.mr);
1851     }
1852 
1853     if (smc->dr_lmb_enabled) {
1854         spapr_create_lmb_dr_connectors(spapr);
1855     }
1856 
1857     filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, "spapr-rtas.bin");
1858     if (!filename) {
1859         error_report("Could not find LPAR rtas '%s'", "spapr-rtas.bin");
1860         exit(1);
1861     }
1862     spapr->rtas_size = get_image_size(filename);
1863     spapr->rtas_blob = g_malloc(spapr->rtas_size);
1864     if (load_image_size(filename, spapr->rtas_blob, spapr->rtas_size) < 0) {
1865         error_report("Could not load LPAR rtas '%s'", filename);
1866         exit(1);
1867     }
1868     if (spapr->rtas_size > RTAS_MAX_SIZE) {
1869         error_report("RTAS too big ! 0x%zx bytes (max is 0x%x)",
1870                      (size_t)spapr->rtas_size, RTAS_MAX_SIZE);
1871         exit(1);
1872     }
1873     g_free(filename);
1874 
1875     /* Set up EPOW events infrastructure */
1876     spapr_events_init(spapr);
1877 
1878     /* Set up the RTC RTAS interfaces */
1879     spapr_rtc_create(spapr);
1880 
1881     /* Set up VIO bus */
1882     spapr->vio_bus = spapr_vio_bus_init();
1883 
1884     for (i = 0; i < MAX_SERIAL_PORTS; i++) {
1885         if (serial_hds[i]) {
1886             spapr_vty_create(spapr->vio_bus, serial_hds[i]);
1887         }
1888     }
1889 
1890     /* We always have at least the nvram device on VIO */
1891     spapr_create_nvram(spapr);
1892 
1893     /* Set up PCI */
1894     spapr_pci_rtas_init();
1895 
1896     phb = spapr_create_phb(spapr, 0);
1897 
1898     for (i = 0; i < nb_nics; i++) {
1899         NICInfo *nd = &nd_table[i];
1900 
1901         if (!nd->model) {
1902             nd->model = g_strdup("ibmveth");
1903         }
1904 
1905         if (strcmp(nd->model, "ibmveth") == 0) {
1906             spapr_vlan_create(spapr->vio_bus, nd);
1907         } else {
1908             pci_nic_init_nofail(&nd_table[i], phb->bus, nd->model, NULL);
1909         }
1910     }
1911 
1912     for (i = 0; i <= drive_get_max_bus(IF_SCSI); i++) {
1913         spapr_vscsi_create(spapr->vio_bus);
1914     }
1915 
1916     /* Graphics */
1917     if (spapr_vga_init(phb->bus)) {
1918         spapr->has_graphics = true;
1919         machine->usb |= defaults_enabled() && !machine->usb_disabled;
1920     }
1921 
1922     if (machine->usb) {
1923         if (smc->use_ohci_by_default) {
1924             pci_create_simple(phb->bus, -1, "pci-ohci");
1925         } else {
1926             pci_create_simple(phb->bus, -1, "nec-usb-xhci");
1927         }
1928 
1929         if (spapr->has_graphics) {
1930             USBBus *usb_bus = usb_bus_find(-1);
1931 
1932             usb_create_simple(usb_bus, "usb-kbd");
1933             usb_create_simple(usb_bus, "usb-mouse");
1934         }
1935     }
1936 
1937     if (spapr->rma_size < (MIN_RMA_SLOF << 20)) {
1938         fprintf(stderr, "qemu: pSeries SLOF firmware requires >= "
1939                 "%ldM guest RMA (Real Mode Area memory)\n", MIN_RMA_SLOF);
1940         exit(1);
1941     }
1942 
1943     if (kernel_filename) {
1944         uint64_t lowaddr = 0;
1945 
1946         kernel_size = load_elf(kernel_filename, translate_kernel_address, NULL,
1947                                NULL, &lowaddr, NULL, 1, PPC_ELF_MACHINE, 0);
1948         if (kernel_size == ELF_LOAD_WRONG_ENDIAN) {
1949             kernel_size = load_elf(kernel_filename,
1950                                    translate_kernel_address, NULL,
1951                                    NULL, &lowaddr, NULL, 0, PPC_ELF_MACHINE, 0);
1952             kernel_le = kernel_size > 0;
1953         }
1954         if (kernel_size < 0) {
1955             fprintf(stderr, "qemu: error loading %s: %s\n",
1956                     kernel_filename, load_elf_strerror(kernel_size));
1957             exit(1);
1958         }
1959 
1960         /* load initrd */
1961         if (initrd_filename) {
1962             /* Try to locate the initrd in the gap between the kernel
1963              * and the firmware. Add a bit of space just in case
1964              */
1965             initrd_base = (KERNEL_LOAD_ADDR + kernel_size + 0x1ffff) & ~0xffff;
1966             initrd_size = load_image_targphys(initrd_filename, initrd_base,
1967                                               load_limit - initrd_base);
1968             if (initrd_size < 0) {
1969                 fprintf(stderr, "qemu: could not load initial ram disk '%s'\n",
1970                         initrd_filename);
1971                 exit(1);
1972             }
1973         } else {
1974             initrd_base = 0;
1975             initrd_size = 0;
1976         }
1977     }
1978 
1979     if (bios_name == NULL) {
1980         bios_name = FW_FILE_NAME;
1981     }
1982     filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, bios_name);
1983     if (!filename) {
1984         error_report("Could not find LPAR firmware '%s'", bios_name);
1985         exit(1);
1986     }
1987     fw_size = load_image_targphys(filename, 0, FW_MAX_SIZE);
1988     if (fw_size <= 0) {
1989         error_report("Could not load LPAR firmware '%s'", filename);
1990         exit(1);
1991     }
1992     g_free(filename);
1993 
1994     /* FIXME: Should register things through the MachineState's qdev
1995      * interface, this is a legacy from the sPAPREnvironment structure
1996      * which predated MachineState but had a similar function */
1997     vmstate_register(NULL, 0, &vmstate_spapr, spapr);
1998     register_savevm_live(NULL, "spapr/htab", -1, 1,
1999                          &savevm_htab_handlers, spapr);
2000 
2001     /* Prepare the device tree */
2002     spapr->fdt_skel = spapr_create_fdt_skel(initrd_base, initrd_size,
2003                                             kernel_size, kernel_le,
2004                                             kernel_cmdline,
2005                                             spapr->check_exception_irq);
2006     assert(spapr->fdt_skel != NULL);
2007 
2008     /* used by RTAS */
2009     QTAILQ_INIT(&spapr->ccs_list);
2010     qemu_register_reset(spapr_ccs_reset_hook, spapr);
2011 
2012     qemu_register_boot_set(spapr_boot_set, spapr);
2013 }
2014 
2015 static int spapr_kvm_type(const char *vm_type)
2016 {
2017     if (!vm_type) {
2018         return 0;
2019     }
2020 
2021     if (!strcmp(vm_type, "HV")) {
2022         return 1;
2023     }
2024 
2025     if (!strcmp(vm_type, "PR")) {
2026         return 2;
2027     }
2028 
2029     error_report("Unknown kvm-type specified '%s'", vm_type);
2030     exit(1);
2031 }
2032 
2033 /*
2034  * Implementation of an interface to adjust firmware path
2035  * for the bootindex property handling.
2036  */
2037 static char *spapr_get_fw_dev_path(FWPathProvider *p, BusState *bus,
2038                                    DeviceState *dev)
2039 {
2040 #define CAST(type, obj, name) \
2041     ((type *)object_dynamic_cast(OBJECT(obj), (name)))
2042     SCSIDevice *d = CAST(SCSIDevice,  dev, TYPE_SCSI_DEVICE);
2043     sPAPRPHBState *phb = CAST(sPAPRPHBState, dev, TYPE_SPAPR_PCI_HOST_BRIDGE);
2044 
2045     if (d) {
2046         void *spapr = CAST(void, bus->parent, "spapr-vscsi");
2047         VirtIOSCSI *virtio = CAST(VirtIOSCSI, bus->parent, TYPE_VIRTIO_SCSI);
2048         USBDevice *usb = CAST(USBDevice, bus->parent, TYPE_USB_DEVICE);
2049 
2050         if (spapr) {
2051             /*
2052              * Replace "channel@0/disk@0,0" with "disk@8000000000000000":
2053              * We use SRP luns of the form 8000 | (bus << 8) | (id << 5) | lun
2054              * in the top 16 bits of the 64-bit LUN
2055              */
2056             unsigned id = 0x8000 | (d->id << 8) | d->lun;
2057             return g_strdup_printf("%s@%"PRIX64, qdev_fw_name(dev),
2058                                    (uint64_t)id << 48);
2059         } else if (virtio) {
2060             /*
2061              * We use SRP luns of the form 01000000 | (target << 8) | lun
2062              * in the top 32 bits of the 64-bit LUN
2063              * Note: the quote above is from SLOF and it is wrong,
2064              * the actual binding is:
2065              * swap 0100 or 10 << or 20 << ( target lun-id -- srplun )
2066              */
2067             unsigned id = 0x1000000 | (d->id << 16) | d->lun;
2068             return g_strdup_printf("%s@%"PRIX64, qdev_fw_name(dev),
2069                                    (uint64_t)id << 32);
2070         } else if (usb) {
2071             /*
2072              * We use SRP luns of the form 01000000 | (usb-port << 16) | lun
2073              * in the top 32 bits of the 64-bit LUN
2074              */
2075             unsigned usb_port = atoi(usb->port->path);
2076             unsigned id = 0x1000000 | (usb_port << 16) | d->lun;
2077             return g_strdup_printf("%s@%"PRIX64, qdev_fw_name(dev),
2078                                    (uint64_t)id << 32);
2079         }
2080     }
2081 
2082     if (phb) {
2083         /* Replace "pci" with "pci@800000020000000" */
2084         return g_strdup_printf("pci@%"PRIX64, phb->buid);
2085     }
2086 
2087     return NULL;
2088 }
2089 
2090 static char *spapr_get_kvm_type(Object *obj, Error **errp)
2091 {
2092     sPAPRMachineState *spapr = SPAPR_MACHINE(obj);
2093 
2094     return g_strdup(spapr->kvm_type);
2095 }
2096 
2097 static void spapr_set_kvm_type(Object *obj, const char *value, Error **errp)
2098 {
2099     sPAPRMachineState *spapr = SPAPR_MACHINE(obj);
2100 
2101     g_free(spapr->kvm_type);
2102     spapr->kvm_type = g_strdup(value);
2103 }
2104 
2105 static void spapr_machine_initfn(Object *obj)
2106 {
2107     object_property_add_str(obj, "kvm-type",
2108                             spapr_get_kvm_type, spapr_set_kvm_type, NULL);
2109     object_property_set_description(obj, "kvm-type",
2110                                     "Specifies the KVM virtualization mode (HV, PR)",
2111                                     NULL);
2112 }
2113 
2114 static void spapr_machine_finalizefn(Object *obj)
2115 {
2116     sPAPRMachineState *spapr = SPAPR_MACHINE(obj);
2117 
2118     g_free(spapr->kvm_type);
2119 }
2120 
2121 static void ppc_cpu_do_nmi_on_cpu(void *arg)
2122 {
2123     CPUState *cs = arg;
2124 
2125     cpu_synchronize_state(cs);
2126     ppc_cpu_do_system_reset(cs);
2127 }
2128 
2129 static void spapr_nmi(NMIState *n, int cpu_index, Error **errp)
2130 {
2131     CPUState *cs;
2132 
2133     CPU_FOREACH(cs) {
2134         async_run_on_cpu(cs, ppc_cpu_do_nmi_on_cpu, cs);
2135     }
2136 }
2137 
2138 static void spapr_add_lmbs(DeviceState *dev, uint64_t addr, uint64_t size,
2139                            uint32_t node, Error **errp)
2140 {
2141     sPAPRDRConnector *drc;
2142     sPAPRDRConnectorClass *drck;
2143     uint32_t nr_lmbs = size/SPAPR_MEMORY_BLOCK_SIZE;
2144     int i, fdt_offset, fdt_size;
2145     void *fdt;
2146 
2147     /*
2148      * Check for DRC connectors and send hotplug notification to the
2149      * guest only in case of hotplugged memory. This allows cold plugged
2150      * memory to be specified at boot time.
2151      */
2152     if (!dev->hotplugged) {
2153         return;
2154     }
2155 
2156     for (i = 0; i < nr_lmbs; i++) {
2157         drc = spapr_dr_connector_by_id(SPAPR_DR_CONNECTOR_TYPE_LMB,
2158                 addr/SPAPR_MEMORY_BLOCK_SIZE);
2159         g_assert(drc);
2160 
2161         fdt = create_device_tree(&fdt_size);
2162         fdt_offset = spapr_populate_memory_node(fdt, node, addr,
2163                                                 SPAPR_MEMORY_BLOCK_SIZE);
2164 
2165         drck = SPAPR_DR_CONNECTOR_GET_CLASS(drc);
2166         drck->attach(drc, dev, fdt, fdt_offset, !dev->hotplugged, errp);
2167         addr += SPAPR_MEMORY_BLOCK_SIZE;
2168     }
2169     spapr_hotplug_req_add_by_count(SPAPR_DR_CONNECTOR_TYPE_LMB, nr_lmbs);
2170 }
2171 
2172 static void spapr_memory_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
2173                               uint32_t node, Error **errp)
2174 {
2175     Error *local_err = NULL;
2176     sPAPRMachineState *ms = SPAPR_MACHINE(hotplug_dev);
2177     PCDIMMDevice *dimm = PC_DIMM(dev);
2178     PCDIMMDeviceClass *ddc = PC_DIMM_GET_CLASS(dimm);
2179     MemoryRegion *mr = ddc->get_memory_region(dimm);
2180     uint64_t align = memory_region_get_alignment(mr);
2181     uint64_t size = memory_region_size(mr);
2182     uint64_t addr;
2183 
2184     if (size % SPAPR_MEMORY_BLOCK_SIZE) {
2185         error_setg(&local_err, "Hotplugged memory size must be a multiple of "
2186                       "%lld MB", SPAPR_MEMORY_BLOCK_SIZE/M_BYTE);
2187         goto out;
2188     }
2189 
2190     pc_dimm_memory_plug(dev, &ms->hotplug_memory, mr, align, &local_err);
2191     if (local_err) {
2192         goto out;
2193     }
2194 
2195     addr = object_property_get_int(OBJECT(dimm), PC_DIMM_ADDR_PROP, &local_err);
2196     if (local_err) {
2197         pc_dimm_memory_unplug(dev, &ms->hotplug_memory, mr);
2198         goto out;
2199     }
2200 
2201     spapr_add_lmbs(dev, addr, size, node, &error_abort);
2202 
2203 out:
2204     error_propagate(errp, local_err);
2205 }
2206 
2207 static void spapr_machine_device_plug(HotplugHandler *hotplug_dev,
2208                                       DeviceState *dev, Error **errp)
2209 {
2210     sPAPRMachineClass *smc = SPAPR_MACHINE_GET_CLASS(qdev_get_machine());
2211 
2212     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2213         int node;
2214 
2215         if (!smc->dr_lmb_enabled) {
2216             error_setg(errp, "Memory hotplug not supported for this machine");
2217             return;
2218         }
2219         node = object_property_get_int(OBJECT(dev), PC_DIMM_NODE_PROP, errp);
2220         if (*errp) {
2221             return;
2222         }
2223 
2224         /*
2225          * Currently PowerPC kernel doesn't allow hot-adding memory to
2226          * memory-less node, but instead will silently add the memory
2227          * to the first node that has some memory. This causes two
2228          * unexpected behaviours for the user.
2229          *
2230          * - Memory gets hotplugged to a different node than what the user
2231          *   specified.
2232          * - Since pc-dimm subsystem in QEMU still thinks that memory belongs
2233          *   to memory-less node, a reboot will set things accordingly
2234          *   and the previously hotplugged memory now ends in the right node.
2235          *   This appears as if some memory moved from one node to another.
2236          *
2237          * So until kernel starts supporting memory hotplug to memory-less
2238          * nodes, just prevent such attempts upfront in QEMU.
2239          */
2240         if (nb_numa_nodes && !numa_info[node].node_mem) {
2241             error_setg(errp, "Can't hotplug memory to memory-less node %d",
2242                        node);
2243             return;
2244         }
2245 
2246         spapr_memory_plug(hotplug_dev, dev, node, errp);
2247     }
2248 }
2249 
2250 static void spapr_machine_device_unplug(HotplugHandler *hotplug_dev,
2251                                       DeviceState *dev, Error **errp)
2252 {
2253     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2254         error_setg(errp, "Memory hot unplug not supported by sPAPR");
2255     }
2256 }
2257 
2258 static HotplugHandler *spapr_get_hotpug_handler(MachineState *machine,
2259                                              DeviceState *dev)
2260 {
2261     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2262         return HOTPLUG_HANDLER(machine);
2263     }
2264     return NULL;
2265 }
2266 
2267 static unsigned spapr_cpu_index_to_socket_id(unsigned cpu_index)
2268 {
2269     /* Allocate to NUMA nodes on a "socket" basis (not that concept of
2270      * socket means much for the paravirtualized PAPR platform) */
2271     return cpu_index / smp_threads / smp_cores;
2272 }
2273 
2274 static void spapr_machine_class_init(ObjectClass *oc, void *data)
2275 {
2276     MachineClass *mc = MACHINE_CLASS(oc);
2277     sPAPRMachineClass *smc = SPAPR_MACHINE_CLASS(oc);
2278     FWPathProviderClass *fwc = FW_PATH_PROVIDER_CLASS(oc);
2279     NMIClass *nc = NMI_CLASS(oc);
2280     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
2281 
2282     mc->desc = "pSeries Logical Partition (PAPR compliant)";
2283 
2284     /*
2285      * We set up the default / latest behaviour here.  The class_init
2286      * functions for the specific versioned machine types can override
2287      * these details for backwards compatibility
2288      */
2289     mc->init = ppc_spapr_init;
2290     mc->reset = ppc_spapr_reset;
2291     mc->block_default_type = IF_SCSI;
2292     mc->max_cpus = MAX_CPUMASK_BITS;
2293     mc->no_parallel = 1;
2294     mc->default_boot_order = "";
2295     mc->default_ram_size = 512 * M_BYTE;
2296     mc->kvm_type = spapr_kvm_type;
2297     mc->has_dynamic_sysbus = true;
2298     mc->pci_allow_0_address = true;
2299     mc->get_hotplug_handler = spapr_get_hotpug_handler;
2300     hc->plug = spapr_machine_device_plug;
2301     hc->unplug = spapr_machine_device_unplug;
2302     mc->cpu_index_to_socket_id = spapr_cpu_index_to_socket_id;
2303 
2304     smc->dr_lmb_enabled = true;
2305     fwc->get_dev_path = spapr_get_fw_dev_path;
2306     nc->nmi_monitor_handler = spapr_nmi;
2307 }
2308 
2309 static const TypeInfo spapr_machine_info = {
2310     .name          = TYPE_SPAPR_MACHINE,
2311     .parent        = TYPE_MACHINE,
2312     .abstract      = true,
2313     .instance_size = sizeof(sPAPRMachineState),
2314     .instance_init = spapr_machine_initfn,
2315     .instance_finalize = spapr_machine_finalizefn,
2316     .class_size    = sizeof(sPAPRMachineClass),
2317     .class_init    = spapr_machine_class_init,
2318     .interfaces = (InterfaceInfo[]) {
2319         { TYPE_FW_PATH_PROVIDER },
2320         { TYPE_NMI },
2321         { TYPE_HOTPLUG_HANDLER },
2322         { }
2323     },
2324 };
2325 
2326 #define DEFINE_SPAPR_MACHINE(suffix, verstr, latest)                 \
2327     static void spapr_machine_##suffix##_class_init(ObjectClass *oc, \
2328                                                     void *data)      \
2329     {                                                                \
2330         MachineClass *mc = MACHINE_CLASS(oc);                        \
2331         spapr_machine_##suffix##_class_options(mc);                  \
2332         if (latest) {                                                \
2333             mc->alias = "pseries";                                   \
2334             mc->is_default = 1;                                      \
2335         }                                                            \
2336     }                                                                \
2337     static void spapr_machine_##suffix##_instance_init(Object *obj)  \
2338     {                                                                \
2339         MachineState *machine = MACHINE(obj);                        \
2340         spapr_machine_##suffix##_instance_options(machine);          \
2341     }                                                                \
2342     static const TypeInfo spapr_machine_##suffix##_info = {          \
2343         .name = MACHINE_TYPE_NAME("pseries-" verstr),                \
2344         .parent = TYPE_SPAPR_MACHINE,                                \
2345         .class_init = spapr_machine_##suffix##_class_init,           \
2346         .instance_init = spapr_machine_##suffix##_instance_init,     \
2347     };                                                               \
2348     static void spapr_machine_register_##suffix(void)                \
2349     {                                                                \
2350         type_register(&spapr_machine_##suffix##_info);               \
2351     }                                                                \
2352     machine_init(spapr_machine_register_##suffix)
2353 
2354 /*
2355  * pseries-2.6
2356  */
2357 static void spapr_machine_2_6_instance_options(MachineState *machine)
2358 {
2359 }
2360 
2361 static void spapr_machine_2_6_class_options(MachineClass *mc)
2362 {
2363     /* Defaults for the latest behaviour inherited from the base class */
2364 }
2365 
2366 DEFINE_SPAPR_MACHINE(2_6, "2.6", true);
2367 
2368 /*
2369  * pseries-2.5
2370  */
2371 #define SPAPR_COMPAT_2_5 \
2372         HW_COMPAT_2_5
2373 
2374 static void spapr_machine_2_5_instance_options(MachineState *machine)
2375 {
2376 }
2377 
2378 static void spapr_machine_2_5_class_options(MachineClass *mc)
2379 {
2380     sPAPRMachineClass *smc = SPAPR_MACHINE_CLASS(mc);
2381 
2382     spapr_machine_2_6_class_options(mc);
2383     smc->use_ohci_by_default = true;
2384     SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_5);
2385 }
2386 
2387 DEFINE_SPAPR_MACHINE(2_5, "2.5", false);
2388 
2389 /*
2390  * pseries-2.4
2391  */
2392 #define SPAPR_COMPAT_2_4 \
2393         HW_COMPAT_2_4
2394 
2395 static void spapr_machine_2_4_instance_options(MachineState *machine)
2396 {
2397     spapr_machine_2_5_instance_options(machine);
2398 }
2399 
2400 static void spapr_machine_2_4_class_options(MachineClass *mc)
2401 {
2402     sPAPRMachineClass *smc = SPAPR_MACHINE_CLASS(mc);
2403 
2404     spapr_machine_2_5_class_options(mc);
2405     smc->dr_lmb_enabled = false;
2406     SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_4);
2407 }
2408 
2409 DEFINE_SPAPR_MACHINE(2_4, "2.4", false);
2410 
2411 /*
2412  * pseries-2.3
2413  */
2414 #define SPAPR_COMPAT_2_3 \
2415         SPAPR_COMPAT_2_4 \
2416         HW_COMPAT_2_3 \
2417         {\
2418             .driver   = "spapr-pci-host-bridge",\
2419             .property = "dynamic-reconfiguration",\
2420             .value    = "off",\
2421         },
2422 
2423 static void spapr_machine_2_3_instance_options(MachineState *machine)
2424 {
2425     spapr_machine_2_4_instance_options(machine);
2426     savevm_skip_section_footers();
2427     global_state_set_optional();
2428 }
2429 
2430 static void spapr_machine_2_3_class_options(MachineClass *mc)
2431 {
2432     spapr_machine_2_4_class_options(mc);
2433     SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_3);
2434 }
2435 DEFINE_SPAPR_MACHINE(2_3, "2.3", false);
2436 
2437 /*
2438  * pseries-2.2
2439  */
2440 
2441 #define SPAPR_COMPAT_2_2 \
2442         SPAPR_COMPAT_2_3 \
2443         HW_COMPAT_2_2 \
2444         {\
2445             .driver   = TYPE_SPAPR_PCI_HOST_BRIDGE,\
2446             .property = "mem_win_size",\
2447             .value    = "0x20000000",\
2448         },
2449 
2450 static void spapr_machine_2_2_instance_options(MachineState *machine)
2451 {
2452     spapr_machine_2_3_instance_options(machine);
2453 }
2454 
2455 static void spapr_machine_2_2_class_options(MachineClass *mc)
2456 {
2457     spapr_machine_2_3_class_options(mc);
2458     SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_2);
2459 }
2460 DEFINE_SPAPR_MACHINE(2_2, "2.2", false);
2461 
2462 /*
2463  * pseries-2.1
2464  */
2465 #define SPAPR_COMPAT_2_1 \
2466         SPAPR_COMPAT_2_2 \
2467         HW_COMPAT_2_1
2468 
2469 static void spapr_machine_2_1_instance_options(MachineState *machine)
2470 {
2471     spapr_machine_2_2_instance_options(machine);
2472 }
2473 
2474 static void spapr_machine_2_1_class_options(MachineClass *mc)
2475 {
2476     spapr_machine_2_2_class_options(mc);
2477     SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_1);
2478 }
2479 DEFINE_SPAPR_MACHINE(2_1, "2.1", false);
2480 
2481 static void spapr_machine_register_types(void)
2482 {
2483     type_register_static(&spapr_machine_info);
2484 }
2485 
2486 type_init(spapr_machine_register_types)
2487