xref: /openbmc/qemu/hw/ppc/spapr.c (revision 0e1d0245)
1 /*
2  * QEMU PowerPC pSeries Logical Partition (aka sPAPR) hardware System Emulator
3  *
4  * Copyright (c) 2004-2007 Fabrice Bellard
5  * Copyright (c) 2007 Jocelyn Mayer
6  * Copyright (c) 2010 David Gibson, IBM Corporation.
7  *
8  * Permission is hereby granted, free of charge, to any person obtaining a copy
9  * of this software and associated documentation files (the "Software"), to deal
10  * in the Software without restriction, including without limitation the rights
11  * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
12  * copies of the Software, and to permit persons to whom the Software is
13  * furnished to do so, subject to the following conditions:
14  *
15  * The above copyright notice and this permission notice shall be included in
16  * all copies or substantial portions of the Software.
17  *
18  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
19  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
20  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
21  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
22  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
23  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
24  * THE SOFTWARE.
25  *
26  */
27 #include "sysemu/sysemu.h"
28 #include "sysemu/numa.h"
29 #include "hw/hw.h"
30 #include "hw/fw-path-provider.h"
31 #include "elf.h"
32 #include "net/net.h"
33 #include "sysemu/device_tree.h"
34 #include "sysemu/block-backend.h"
35 #include "sysemu/cpus.h"
36 #include "sysemu/kvm.h"
37 #include "sysemu/device_tree.h"
38 #include "kvm_ppc.h"
39 #include "migration/migration.h"
40 #include "mmu-hash64.h"
41 #include "qom/cpu.h"
42 
43 #include "hw/boards.h"
44 #include "hw/ppc/ppc.h"
45 #include "hw/loader.h"
46 
47 #include "hw/ppc/spapr.h"
48 #include "hw/ppc/spapr_vio.h"
49 #include "hw/pci-host/spapr.h"
50 #include "hw/ppc/xics.h"
51 #include "hw/pci/msi.h"
52 
53 #include "hw/pci/pci.h"
54 #include "hw/scsi/scsi.h"
55 #include "hw/virtio/virtio-scsi.h"
56 
57 #include "exec/address-spaces.h"
58 #include "hw/usb.h"
59 #include "qemu/config-file.h"
60 #include "qemu/error-report.h"
61 #include "trace.h"
62 #include "hw/nmi.h"
63 
64 #include "hw/compat.h"
65 #include "qemu-common.h"
66 
67 #include <libfdt.h>
68 
69 /* SLOF memory layout:
70  *
71  * SLOF raw image loaded at 0, copies its romfs right below the flat
72  * device-tree, then position SLOF itself 31M below that
73  *
74  * So we set FW_OVERHEAD to 40MB which should account for all of that
75  * and more
76  *
77  * We load our kernel at 4M, leaving space for SLOF initial image
78  */
79 #define FDT_MAX_SIZE            0x100000
80 #define RTAS_MAX_SIZE           0x10000
81 #define RTAS_MAX_ADDR           0x80000000 /* RTAS must stay below that */
82 #define FW_MAX_SIZE             0x400000
83 #define FW_FILE_NAME            "slof.bin"
84 #define FW_OVERHEAD             0x2800000
85 #define KERNEL_LOAD_ADDR        FW_MAX_SIZE
86 
87 #define MIN_RMA_SLOF            128UL
88 
89 #define TIMEBASE_FREQ           512000000ULL
90 
91 #define PHANDLE_XICP            0x00001111
92 
93 #define HTAB_SIZE(spapr)        (1ULL << ((spapr)->htab_shift))
94 
95 static XICSState *try_create_xics(const char *type, int nr_servers,
96                                   int nr_irqs, Error **errp)
97 {
98     Error *err = NULL;
99     DeviceState *dev;
100 
101     dev = qdev_create(NULL, type);
102     qdev_prop_set_uint32(dev, "nr_servers", nr_servers);
103     qdev_prop_set_uint32(dev, "nr_irqs", nr_irqs);
104     object_property_set_bool(OBJECT(dev), true, "realized", &err);
105     if (err) {
106         error_propagate(errp, err);
107         object_unparent(OBJECT(dev));
108         return NULL;
109     }
110     return XICS_COMMON(dev);
111 }
112 
113 static XICSState *xics_system_init(MachineState *machine,
114                                    int nr_servers, int nr_irqs)
115 {
116     XICSState *icp = NULL;
117 
118     if (kvm_enabled()) {
119         Error *err = NULL;
120 
121         if (machine_kernel_irqchip_allowed(machine)) {
122             icp = try_create_xics(TYPE_KVM_XICS, nr_servers, nr_irqs, &err);
123         }
124         if (machine_kernel_irqchip_required(machine) && !icp) {
125             error_report("kernel_irqchip requested but unavailable: %s",
126                          error_get_pretty(err));
127         }
128         error_free(err);
129     }
130 
131     if (!icp) {
132         icp = try_create_xics(TYPE_XICS, nr_servers, nr_irqs, &error_abort);
133     }
134 
135     return icp;
136 }
137 
138 static int spapr_fixup_cpu_smt_dt(void *fdt, int offset, PowerPCCPU *cpu,
139                                   int smt_threads)
140 {
141     int i, ret = 0;
142     uint32_t servers_prop[smt_threads];
143     uint32_t gservers_prop[smt_threads * 2];
144     int index = ppc_get_vcpu_dt_id(cpu);
145 
146     if (cpu->cpu_version) {
147         ret = fdt_setprop_cell(fdt, offset, "cpu-version", cpu->cpu_version);
148         if (ret < 0) {
149             return ret;
150         }
151     }
152 
153     /* Build interrupt servers and gservers properties */
154     for (i = 0; i < smt_threads; i++) {
155         servers_prop[i] = cpu_to_be32(index + i);
156         /* Hack, direct the group queues back to cpu 0 */
157         gservers_prop[i*2] = cpu_to_be32(index + i);
158         gservers_prop[i*2 + 1] = 0;
159     }
160     ret = fdt_setprop(fdt, offset, "ibm,ppc-interrupt-server#s",
161                       servers_prop, sizeof(servers_prop));
162     if (ret < 0) {
163         return ret;
164     }
165     ret = fdt_setprop(fdt, offset, "ibm,ppc-interrupt-gserver#s",
166                       gservers_prop, sizeof(gservers_prop));
167 
168     return ret;
169 }
170 
171 static int spapr_fixup_cpu_numa_dt(void *fdt, int offset, CPUState *cs)
172 {
173     int ret = 0;
174     PowerPCCPU *cpu = POWERPC_CPU(cs);
175     int index = ppc_get_vcpu_dt_id(cpu);
176     uint32_t associativity[] = {cpu_to_be32(0x5),
177                                 cpu_to_be32(0x0),
178                                 cpu_to_be32(0x0),
179                                 cpu_to_be32(0x0),
180                                 cpu_to_be32(cs->numa_node),
181                                 cpu_to_be32(index)};
182 
183     /* Advertise NUMA via ibm,associativity */
184     if (nb_numa_nodes > 1) {
185         ret = fdt_setprop(fdt, offset, "ibm,associativity", associativity,
186                           sizeof(associativity));
187     }
188 
189     return ret;
190 }
191 
192 static int spapr_fixup_cpu_dt(void *fdt, sPAPRMachineState *spapr)
193 {
194     int ret = 0, offset, cpus_offset;
195     CPUState *cs;
196     char cpu_model[32];
197     int smt = kvmppc_smt_threads();
198     uint32_t pft_size_prop[] = {0, cpu_to_be32(spapr->htab_shift)};
199 
200     CPU_FOREACH(cs) {
201         PowerPCCPU *cpu = POWERPC_CPU(cs);
202         DeviceClass *dc = DEVICE_GET_CLASS(cs);
203         int index = ppc_get_vcpu_dt_id(cpu);
204 
205         if ((index % smt) != 0) {
206             continue;
207         }
208 
209         snprintf(cpu_model, 32, "%s@%x", dc->fw_name, index);
210 
211         cpus_offset = fdt_path_offset(fdt, "/cpus");
212         if (cpus_offset < 0) {
213             cpus_offset = fdt_add_subnode(fdt, fdt_path_offset(fdt, "/"),
214                                           "cpus");
215             if (cpus_offset < 0) {
216                 return cpus_offset;
217             }
218         }
219         offset = fdt_subnode_offset(fdt, cpus_offset, cpu_model);
220         if (offset < 0) {
221             offset = fdt_add_subnode(fdt, cpus_offset, cpu_model);
222             if (offset < 0) {
223                 return offset;
224             }
225         }
226 
227         ret = fdt_setprop(fdt, offset, "ibm,pft-size",
228                           pft_size_prop, sizeof(pft_size_prop));
229         if (ret < 0) {
230             return ret;
231         }
232 
233         ret = spapr_fixup_cpu_numa_dt(fdt, offset, cs);
234         if (ret < 0) {
235             return ret;
236         }
237 
238         ret = spapr_fixup_cpu_smt_dt(fdt, offset, cpu,
239                                      ppc_get_compat_smt_threads(cpu));
240         if (ret < 0) {
241             return ret;
242         }
243     }
244     return ret;
245 }
246 
247 
248 static size_t create_page_sizes_prop(CPUPPCState *env, uint32_t *prop,
249                                      size_t maxsize)
250 {
251     size_t maxcells = maxsize / sizeof(uint32_t);
252     int i, j, count;
253     uint32_t *p = prop;
254 
255     for (i = 0; i < PPC_PAGE_SIZES_MAX_SZ; i++) {
256         struct ppc_one_seg_page_size *sps = &env->sps.sps[i];
257 
258         if (!sps->page_shift) {
259             break;
260         }
261         for (count = 0; count < PPC_PAGE_SIZES_MAX_SZ; count++) {
262             if (sps->enc[count].page_shift == 0) {
263                 break;
264             }
265         }
266         if ((p - prop) >= (maxcells - 3 - count * 2)) {
267             break;
268         }
269         *(p++) = cpu_to_be32(sps->page_shift);
270         *(p++) = cpu_to_be32(sps->slb_enc);
271         *(p++) = cpu_to_be32(count);
272         for (j = 0; j < count; j++) {
273             *(p++) = cpu_to_be32(sps->enc[j].page_shift);
274             *(p++) = cpu_to_be32(sps->enc[j].pte_enc);
275         }
276     }
277 
278     return (p - prop) * sizeof(uint32_t);
279 }
280 
281 static hwaddr spapr_node0_size(void)
282 {
283     MachineState *machine = MACHINE(qdev_get_machine());
284 
285     if (nb_numa_nodes) {
286         int i;
287         for (i = 0; i < nb_numa_nodes; ++i) {
288             if (numa_info[i].node_mem) {
289                 return MIN(pow2floor(numa_info[i].node_mem),
290                            machine->ram_size);
291             }
292         }
293     }
294     return machine->ram_size;
295 }
296 
297 #define _FDT(exp) \
298     do { \
299         int ret = (exp);                                           \
300         if (ret < 0) {                                             \
301             fprintf(stderr, "qemu: error creating device tree: %s: %s\n", \
302                     #exp, fdt_strerror(ret));                      \
303             exit(1);                                               \
304         }                                                          \
305     } while (0)
306 
307 static void add_str(GString *s, const gchar *s1)
308 {
309     g_string_append_len(s, s1, strlen(s1) + 1);
310 }
311 
312 static void *spapr_create_fdt_skel(hwaddr initrd_base,
313                                    hwaddr initrd_size,
314                                    hwaddr kernel_size,
315                                    bool little_endian,
316                                    const char *kernel_cmdline,
317                                    uint32_t epow_irq)
318 {
319     void *fdt;
320     uint32_t start_prop = cpu_to_be32(initrd_base);
321     uint32_t end_prop = cpu_to_be32(initrd_base + initrd_size);
322     GString *hypertas = g_string_sized_new(256);
323     GString *qemu_hypertas = g_string_sized_new(256);
324     uint32_t refpoints[] = {cpu_to_be32(0x4), cpu_to_be32(0x4)};
325     uint32_t interrupt_server_ranges_prop[] = {0, cpu_to_be32(max_cpus)};
326     unsigned char vec5[] = {0x0, 0x0, 0x0, 0x0, 0x0, 0x80};
327     char *buf;
328 
329     add_str(hypertas, "hcall-pft");
330     add_str(hypertas, "hcall-term");
331     add_str(hypertas, "hcall-dabr");
332     add_str(hypertas, "hcall-interrupt");
333     add_str(hypertas, "hcall-tce");
334     add_str(hypertas, "hcall-vio");
335     add_str(hypertas, "hcall-splpar");
336     add_str(hypertas, "hcall-bulk");
337     add_str(hypertas, "hcall-set-mode");
338     add_str(qemu_hypertas, "hcall-memop1");
339 
340     fdt = g_malloc0(FDT_MAX_SIZE);
341     _FDT((fdt_create(fdt, FDT_MAX_SIZE)));
342 
343     if (kernel_size) {
344         _FDT((fdt_add_reservemap_entry(fdt, KERNEL_LOAD_ADDR, kernel_size)));
345     }
346     if (initrd_size) {
347         _FDT((fdt_add_reservemap_entry(fdt, initrd_base, initrd_size)));
348     }
349     _FDT((fdt_finish_reservemap(fdt)));
350 
351     /* Root node */
352     _FDT((fdt_begin_node(fdt, "")));
353     _FDT((fdt_property_string(fdt, "device_type", "chrp")));
354     _FDT((fdt_property_string(fdt, "model", "IBM pSeries (emulated by qemu)")));
355     _FDT((fdt_property_string(fdt, "compatible", "qemu,pseries")));
356 
357     /*
358      * Add info to guest to indentify which host is it being run on
359      * and what is the uuid of the guest
360      */
361     if (kvmppc_get_host_model(&buf)) {
362         _FDT((fdt_property_string(fdt, "host-model", buf)));
363         g_free(buf);
364     }
365     if (kvmppc_get_host_serial(&buf)) {
366         _FDT((fdt_property_string(fdt, "host-serial", buf)));
367         g_free(buf);
368     }
369 
370     buf = g_strdup_printf(UUID_FMT, qemu_uuid[0], qemu_uuid[1],
371                           qemu_uuid[2], qemu_uuid[3], qemu_uuid[4],
372                           qemu_uuid[5], qemu_uuid[6], qemu_uuid[7],
373                           qemu_uuid[8], qemu_uuid[9], qemu_uuid[10],
374                           qemu_uuid[11], qemu_uuid[12], qemu_uuid[13],
375                           qemu_uuid[14], qemu_uuid[15]);
376 
377     _FDT((fdt_property_string(fdt, "vm,uuid", buf)));
378     g_free(buf);
379 
380     if (qemu_get_vm_name()) {
381         _FDT((fdt_property_string(fdt, "ibm,partition-name",
382                                   qemu_get_vm_name())));
383     }
384 
385     _FDT((fdt_property_cell(fdt, "#address-cells", 0x2)));
386     _FDT((fdt_property_cell(fdt, "#size-cells", 0x2)));
387 
388     /* /chosen */
389     _FDT((fdt_begin_node(fdt, "chosen")));
390 
391     /* Set Form1_affinity */
392     _FDT((fdt_property(fdt, "ibm,architecture-vec-5", vec5, sizeof(vec5))));
393 
394     _FDT((fdt_property_string(fdt, "bootargs", kernel_cmdline)));
395     _FDT((fdt_property(fdt, "linux,initrd-start",
396                        &start_prop, sizeof(start_prop))));
397     _FDT((fdt_property(fdt, "linux,initrd-end",
398                        &end_prop, sizeof(end_prop))));
399     if (kernel_size) {
400         uint64_t kprop[2] = { cpu_to_be64(KERNEL_LOAD_ADDR),
401                               cpu_to_be64(kernel_size) };
402 
403         _FDT((fdt_property(fdt, "qemu,boot-kernel", &kprop, sizeof(kprop))));
404         if (little_endian) {
405             _FDT((fdt_property(fdt, "qemu,boot-kernel-le", NULL, 0)));
406         }
407     }
408     if (boot_menu) {
409         _FDT((fdt_property_cell(fdt, "qemu,boot-menu", boot_menu)));
410     }
411     _FDT((fdt_property_cell(fdt, "qemu,graphic-width", graphic_width)));
412     _FDT((fdt_property_cell(fdt, "qemu,graphic-height", graphic_height)));
413     _FDT((fdt_property_cell(fdt, "qemu,graphic-depth", graphic_depth)));
414 
415     _FDT((fdt_end_node(fdt)));
416 
417     /* RTAS */
418     _FDT((fdt_begin_node(fdt, "rtas")));
419 
420     if (!kvm_enabled() || kvmppc_spapr_use_multitce()) {
421         add_str(hypertas, "hcall-multi-tce");
422     }
423     _FDT((fdt_property(fdt, "ibm,hypertas-functions", hypertas->str,
424                        hypertas->len)));
425     g_string_free(hypertas, TRUE);
426     _FDT((fdt_property(fdt, "qemu,hypertas-functions", qemu_hypertas->str,
427                        qemu_hypertas->len)));
428     g_string_free(qemu_hypertas, TRUE);
429 
430     _FDT((fdt_property(fdt, "ibm,associativity-reference-points",
431         refpoints, sizeof(refpoints))));
432 
433     _FDT((fdt_property_cell(fdt, "rtas-error-log-max", RTAS_ERROR_LOG_MAX)));
434     _FDT((fdt_property_cell(fdt, "rtas-event-scan-rate",
435                             RTAS_EVENT_SCAN_RATE)));
436 
437     if (msi_supported) {
438         _FDT((fdt_property(fdt, "ibm,change-msix-capable", NULL, 0)));
439     }
440 
441     /*
442      * According to PAPR, rtas ibm,os-term does not guarantee a return
443      * back to the guest cpu.
444      *
445      * While an additional ibm,extended-os-term property indicates that
446      * rtas call return will always occur. Set this property.
447      */
448     _FDT((fdt_property(fdt, "ibm,extended-os-term", NULL, 0)));
449 
450     _FDT((fdt_end_node(fdt)));
451 
452     /* interrupt controller */
453     _FDT((fdt_begin_node(fdt, "interrupt-controller")));
454 
455     _FDT((fdt_property_string(fdt, "device_type",
456                               "PowerPC-External-Interrupt-Presentation")));
457     _FDT((fdt_property_string(fdt, "compatible", "IBM,ppc-xicp")));
458     _FDT((fdt_property(fdt, "interrupt-controller", NULL, 0)));
459     _FDT((fdt_property(fdt, "ibm,interrupt-server-ranges",
460                        interrupt_server_ranges_prop,
461                        sizeof(interrupt_server_ranges_prop))));
462     _FDT((fdt_property_cell(fdt, "#interrupt-cells", 2)));
463     _FDT((fdt_property_cell(fdt, "linux,phandle", PHANDLE_XICP)));
464     _FDT((fdt_property_cell(fdt, "phandle", PHANDLE_XICP)));
465 
466     _FDT((fdt_end_node(fdt)));
467 
468     /* vdevice */
469     _FDT((fdt_begin_node(fdt, "vdevice")));
470 
471     _FDT((fdt_property_string(fdt, "device_type", "vdevice")));
472     _FDT((fdt_property_string(fdt, "compatible", "IBM,vdevice")));
473     _FDT((fdt_property_cell(fdt, "#address-cells", 0x1)));
474     _FDT((fdt_property_cell(fdt, "#size-cells", 0x0)));
475     _FDT((fdt_property_cell(fdt, "#interrupt-cells", 0x2)));
476     _FDT((fdt_property(fdt, "interrupt-controller", NULL, 0)));
477 
478     _FDT((fdt_end_node(fdt)));
479 
480     /* event-sources */
481     spapr_events_fdt_skel(fdt, epow_irq);
482 
483     /* /hypervisor node */
484     if (kvm_enabled()) {
485         uint8_t hypercall[16];
486 
487         /* indicate KVM hypercall interface */
488         _FDT((fdt_begin_node(fdt, "hypervisor")));
489         _FDT((fdt_property_string(fdt, "compatible", "linux,kvm")));
490         if (kvmppc_has_cap_fixup_hcalls()) {
491             /*
492              * Older KVM versions with older guest kernels were broken with the
493              * magic page, don't allow the guest to map it.
494              */
495             kvmppc_get_hypercall(first_cpu->env_ptr, hypercall,
496                                  sizeof(hypercall));
497             _FDT((fdt_property(fdt, "hcall-instructions", hypercall,
498                               sizeof(hypercall))));
499         }
500         _FDT((fdt_end_node(fdt)));
501     }
502 
503     _FDT((fdt_end_node(fdt))); /* close root node */
504     _FDT((fdt_finish(fdt)));
505 
506     return fdt;
507 }
508 
509 static int spapr_populate_memory_node(void *fdt, int nodeid, hwaddr start,
510                                        hwaddr size)
511 {
512     uint32_t associativity[] = {
513         cpu_to_be32(0x4), /* length */
514         cpu_to_be32(0x0), cpu_to_be32(0x0),
515         cpu_to_be32(0x0), cpu_to_be32(nodeid)
516     };
517     char mem_name[32];
518     uint64_t mem_reg_property[2];
519     int off;
520 
521     mem_reg_property[0] = cpu_to_be64(start);
522     mem_reg_property[1] = cpu_to_be64(size);
523 
524     sprintf(mem_name, "memory@" TARGET_FMT_lx, start);
525     off = fdt_add_subnode(fdt, 0, mem_name);
526     _FDT(off);
527     _FDT((fdt_setprop_string(fdt, off, "device_type", "memory")));
528     _FDT((fdt_setprop(fdt, off, "reg", mem_reg_property,
529                       sizeof(mem_reg_property))));
530     _FDT((fdt_setprop(fdt, off, "ibm,associativity", associativity,
531                       sizeof(associativity))));
532     return off;
533 }
534 
535 static int spapr_populate_memory(sPAPRMachineState *spapr, void *fdt)
536 {
537     MachineState *machine = MACHINE(spapr);
538     hwaddr mem_start, node_size;
539     int i, nb_nodes = nb_numa_nodes;
540     NodeInfo *nodes = numa_info;
541     NodeInfo ramnode;
542 
543     /* No NUMA nodes, assume there is just one node with whole RAM */
544     if (!nb_numa_nodes) {
545         nb_nodes = 1;
546         ramnode.node_mem = machine->ram_size;
547         nodes = &ramnode;
548     }
549 
550     for (i = 0, mem_start = 0; i < nb_nodes; ++i) {
551         if (!nodes[i].node_mem) {
552             continue;
553         }
554         if (mem_start >= machine->ram_size) {
555             node_size = 0;
556         } else {
557             node_size = nodes[i].node_mem;
558             if (node_size > machine->ram_size - mem_start) {
559                 node_size = machine->ram_size - mem_start;
560             }
561         }
562         if (!mem_start) {
563             /* ppc_spapr_init() checks for rma_size <= node0_size already */
564             spapr_populate_memory_node(fdt, i, 0, spapr->rma_size);
565             mem_start += spapr->rma_size;
566             node_size -= spapr->rma_size;
567         }
568         for ( ; node_size; ) {
569             hwaddr sizetmp = pow2floor(node_size);
570 
571             /* mem_start != 0 here */
572             if (ctzl(mem_start) < ctzl(sizetmp)) {
573                 sizetmp = 1ULL << ctzl(mem_start);
574             }
575 
576             spapr_populate_memory_node(fdt, i, mem_start, sizetmp);
577             node_size -= sizetmp;
578             mem_start += sizetmp;
579         }
580     }
581 
582     return 0;
583 }
584 
585 static void spapr_populate_cpu_dt(CPUState *cs, void *fdt, int offset,
586                                   sPAPRMachineState *spapr)
587 {
588     PowerPCCPU *cpu = POWERPC_CPU(cs);
589     CPUPPCState *env = &cpu->env;
590     PowerPCCPUClass *pcc = POWERPC_CPU_GET_CLASS(cs);
591     int index = ppc_get_vcpu_dt_id(cpu);
592     uint32_t segs[] = {cpu_to_be32(28), cpu_to_be32(40),
593                        0xffffffff, 0xffffffff};
594     uint32_t tbfreq = kvm_enabled() ? kvmppc_get_tbfreq() : TIMEBASE_FREQ;
595     uint32_t cpufreq = kvm_enabled() ? kvmppc_get_clockfreq() : 1000000000;
596     uint32_t page_sizes_prop[64];
597     size_t page_sizes_prop_size;
598     uint32_t vcpus_per_socket = smp_threads * smp_cores;
599     uint32_t pft_size_prop[] = {0, cpu_to_be32(spapr->htab_shift)};
600 
601     /* Note: we keep CI large pages off for now because a 64K capable guest
602      * provisioned with large pages might otherwise try to map a qemu
603      * framebuffer (or other kind of memory mapped PCI BAR) using 64K pages
604      * even if that qemu runs on a 4k host.
605      *
606      * We can later add this bit back when we are confident this is not
607      * an issue (!HV KVM or 64K host)
608      */
609     uint8_t pa_features_206[] = { 6, 0,
610         0xf6, 0x1f, 0xc7, 0x00, 0x80, 0xc0 };
611     uint8_t pa_features_207[] = { 24, 0,
612         0xf6, 0x1f, 0xc7, 0xc0, 0x80, 0xf0,
613         0x80, 0x00, 0x00, 0x00, 0x00, 0x00,
614         0x00, 0x00, 0x00, 0x00, 0x80, 0x00,
615         0x80, 0x00, 0x80, 0x00, 0x80, 0x00 };
616     uint8_t *pa_features;
617     size_t pa_size;
618 
619     _FDT((fdt_setprop_cell(fdt, offset, "reg", index)));
620     _FDT((fdt_setprop_string(fdt, offset, "device_type", "cpu")));
621 
622     _FDT((fdt_setprop_cell(fdt, offset, "cpu-version", env->spr[SPR_PVR])));
623     _FDT((fdt_setprop_cell(fdt, offset, "d-cache-block-size",
624                            env->dcache_line_size)));
625     _FDT((fdt_setprop_cell(fdt, offset, "d-cache-line-size",
626                            env->dcache_line_size)));
627     _FDT((fdt_setprop_cell(fdt, offset, "i-cache-block-size",
628                            env->icache_line_size)));
629     _FDT((fdt_setprop_cell(fdt, offset, "i-cache-line-size",
630                            env->icache_line_size)));
631 
632     if (pcc->l1_dcache_size) {
633         _FDT((fdt_setprop_cell(fdt, offset, "d-cache-size",
634                                pcc->l1_dcache_size)));
635     } else {
636         fprintf(stderr, "Warning: Unknown L1 dcache size for cpu\n");
637     }
638     if (pcc->l1_icache_size) {
639         _FDT((fdt_setprop_cell(fdt, offset, "i-cache-size",
640                                pcc->l1_icache_size)));
641     } else {
642         fprintf(stderr, "Warning: Unknown L1 icache size for cpu\n");
643     }
644 
645     _FDT((fdt_setprop_cell(fdt, offset, "timebase-frequency", tbfreq)));
646     _FDT((fdt_setprop_cell(fdt, offset, "clock-frequency", cpufreq)));
647     _FDT((fdt_setprop_cell(fdt, offset, "slb-size", env->slb_nr)));
648     _FDT((fdt_setprop_cell(fdt, offset, "ibm,slb-size", env->slb_nr)));
649     _FDT((fdt_setprop_string(fdt, offset, "status", "okay")));
650     _FDT((fdt_setprop(fdt, offset, "64-bit", NULL, 0)));
651 
652     if (env->spr_cb[SPR_PURR].oea_read) {
653         _FDT((fdt_setprop(fdt, offset, "ibm,purr", NULL, 0)));
654     }
655 
656     if (env->mmu_model & POWERPC_MMU_1TSEG) {
657         _FDT((fdt_setprop(fdt, offset, "ibm,processor-segment-sizes",
658                           segs, sizeof(segs))));
659     }
660 
661     /* Advertise VMX/VSX (vector extensions) if available
662      *   0 / no property == no vector extensions
663      *   1               == VMX / Altivec available
664      *   2               == VSX available */
665     if (env->insns_flags & PPC_ALTIVEC) {
666         uint32_t vmx = (env->insns_flags2 & PPC2_VSX) ? 2 : 1;
667 
668         _FDT((fdt_setprop_cell(fdt, offset, "ibm,vmx", vmx)));
669     }
670 
671     /* Advertise DFP (Decimal Floating Point) if available
672      *   0 / no property == no DFP
673      *   1               == DFP available */
674     if (env->insns_flags2 & PPC2_DFP) {
675         _FDT((fdt_setprop_cell(fdt, offset, "ibm,dfp", 1)));
676     }
677 
678     page_sizes_prop_size = create_page_sizes_prop(env, page_sizes_prop,
679                                                   sizeof(page_sizes_prop));
680     if (page_sizes_prop_size) {
681         _FDT((fdt_setprop(fdt, offset, "ibm,segment-page-sizes",
682                           page_sizes_prop, page_sizes_prop_size)));
683     }
684 
685     /* Do the ibm,pa-features property, adjust it for ci-large-pages */
686     if (env->mmu_model == POWERPC_MMU_2_06) {
687         pa_features = pa_features_206;
688         pa_size = sizeof(pa_features_206);
689     } else /* env->mmu_model == POWERPC_MMU_2_07 */ {
690         pa_features = pa_features_207;
691         pa_size = sizeof(pa_features_207);
692     }
693     if (env->ci_large_pages) {
694         pa_features[3] |= 0x20;
695     }
696     _FDT((fdt_setprop(fdt, offset, "ibm,pa-features", pa_features, pa_size)));
697 
698     _FDT((fdt_setprop_cell(fdt, offset, "ibm,chip-id",
699                            cs->cpu_index / vcpus_per_socket)));
700 
701     _FDT((fdt_setprop(fdt, offset, "ibm,pft-size",
702                       pft_size_prop, sizeof(pft_size_prop))));
703 
704     _FDT(spapr_fixup_cpu_numa_dt(fdt, offset, cs));
705 
706     _FDT(spapr_fixup_cpu_smt_dt(fdt, offset, cpu,
707                                 ppc_get_compat_smt_threads(cpu)));
708 }
709 
710 static void spapr_populate_cpus_dt_node(void *fdt, sPAPRMachineState *spapr)
711 {
712     CPUState *cs;
713     int cpus_offset;
714     char *nodename;
715     int smt = kvmppc_smt_threads();
716 
717     cpus_offset = fdt_add_subnode(fdt, 0, "cpus");
718     _FDT(cpus_offset);
719     _FDT((fdt_setprop_cell(fdt, cpus_offset, "#address-cells", 0x1)));
720     _FDT((fdt_setprop_cell(fdt, cpus_offset, "#size-cells", 0x0)));
721 
722     /*
723      * We walk the CPUs in reverse order to ensure that CPU DT nodes
724      * created by fdt_add_subnode() end up in the right order in FDT
725      * for the guest kernel the enumerate the CPUs correctly.
726      */
727     CPU_FOREACH_REVERSE(cs) {
728         PowerPCCPU *cpu = POWERPC_CPU(cs);
729         int index = ppc_get_vcpu_dt_id(cpu);
730         DeviceClass *dc = DEVICE_GET_CLASS(cs);
731         int offset;
732 
733         if ((index % smt) != 0) {
734             continue;
735         }
736 
737         nodename = g_strdup_printf("%s@%x", dc->fw_name, index);
738         offset = fdt_add_subnode(fdt, cpus_offset, nodename);
739         g_free(nodename);
740         _FDT(offset);
741         spapr_populate_cpu_dt(cs, fdt, offset, spapr);
742     }
743 
744 }
745 
746 /*
747  * Adds ibm,dynamic-reconfiguration-memory node.
748  * Refer to docs/specs/ppc-spapr-hotplug.txt for the documentation
749  * of this device tree node.
750  */
751 static int spapr_populate_drconf_memory(sPAPRMachineState *spapr, void *fdt)
752 {
753     MachineState *machine = MACHINE(spapr);
754     int ret, i, offset;
755     uint64_t lmb_size = SPAPR_MEMORY_BLOCK_SIZE;
756     uint32_t prop_lmb_size[] = {0, cpu_to_be32(lmb_size)};
757     uint32_t nr_lmbs = (machine->maxram_size - machine->ram_size)/lmb_size;
758     uint32_t *int_buf, *cur_index, buf_len;
759     int nr_nodes = nb_numa_nodes ? nb_numa_nodes : 1;
760 
761     /*
762      * Allocate enough buffer size to fit in ibm,dynamic-memory
763      * or ibm,associativity-lookup-arrays
764      */
765     buf_len = MAX(nr_lmbs * SPAPR_DR_LMB_LIST_ENTRY_SIZE + 1, nr_nodes * 4 + 2)
766               * sizeof(uint32_t);
767     cur_index = int_buf = g_malloc0(buf_len);
768 
769     offset = fdt_add_subnode(fdt, 0, "ibm,dynamic-reconfiguration-memory");
770 
771     ret = fdt_setprop(fdt, offset, "ibm,lmb-size", prop_lmb_size,
772                     sizeof(prop_lmb_size));
773     if (ret < 0) {
774         goto out;
775     }
776 
777     ret = fdt_setprop_cell(fdt, offset, "ibm,memory-flags-mask", 0xff);
778     if (ret < 0) {
779         goto out;
780     }
781 
782     ret = fdt_setprop_cell(fdt, offset, "ibm,memory-preservation-time", 0x0);
783     if (ret < 0) {
784         goto out;
785     }
786 
787     /* ibm,dynamic-memory */
788     int_buf[0] = cpu_to_be32(nr_lmbs);
789     cur_index++;
790     for (i = 0; i < nr_lmbs; i++) {
791         sPAPRDRConnector *drc;
792         sPAPRDRConnectorClass *drck;
793         uint64_t addr = i * lmb_size + spapr->hotplug_memory.base;;
794         uint32_t *dynamic_memory = cur_index;
795 
796         drc = spapr_dr_connector_by_id(SPAPR_DR_CONNECTOR_TYPE_LMB,
797                                        addr/lmb_size);
798         g_assert(drc);
799         drck = SPAPR_DR_CONNECTOR_GET_CLASS(drc);
800 
801         dynamic_memory[0] = cpu_to_be32(addr >> 32);
802         dynamic_memory[1] = cpu_to_be32(addr & 0xffffffff);
803         dynamic_memory[2] = cpu_to_be32(drck->get_index(drc));
804         dynamic_memory[3] = cpu_to_be32(0); /* reserved */
805         dynamic_memory[4] = cpu_to_be32(numa_get_node(addr, NULL));
806         if (addr < machine->ram_size ||
807                     memory_region_present(get_system_memory(), addr)) {
808             dynamic_memory[5] = cpu_to_be32(SPAPR_LMB_FLAGS_ASSIGNED);
809         } else {
810             dynamic_memory[5] = cpu_to_be32(0);
811         }
812 
813         cur_index += SPAPR_DR_LMB_LIST_ENTRY_SIZE;
814     }
815     ret = fdt_setprop(fdt, offset, "ibm,dynamic-memory", int_buf, buf_len);
816     if (ret < 0) {
817         goto out;
818     }
819 
820     /* ibm,associativity-lookup-arrays */
821     cur_index = int_buf;
822     int_buf[0] = cpu_to_be32(nr_nodes);
823     int_buf[1] = cpu_to_be32(4); /* Number of entries per associativity list */
824     cur_index += 2;
825     for (i = 0; i < nr_nodes; i++) {
826         uint32_t associativity[] = {
827             cpu_to_be32(0x0),
828             cpu_to_be32(0x0),
829             cpu_to_be32(0x0),
830             cpu_to_be32(i)
831         };
832         memcpy(cur_index, associativity, sizeof(associativity));
833         cur_index += 4;
834     }
835     ret = fdt_setprop(fdt, offset, "ibm,associativity-lookup-arrays", int_buf,
836             (cur_index - int_buf) * sizeof(uint32_t));
837 out:
838     g_free(int_buf);
839     return ret;
840 }
841 
842 int spapr_h_cas_compose_response(sPAPRMachineState *spapr,
843                                  target_ulong addr, target_ulong size,
844                                  bool cpu_update, bool memory_update)
845 {
846     void *fdt, *fdt_skel;
847     sPAPRDeviceTreeUpdateHeader hdr = { .version_id = 1 };
848     sPAPRMachineClass *smc = SPAPR_MACHINE_GET_CLASS(qdev_get_machine());
849 
850     size -= sizeof(hdr);
851 
852     /* Create sceleton */
853     fdt_skel = g_malloc0(size);
854     _FDT((fdt_create(fdt_skel, size)));
855     _FDT((fdt_begin_node(fdt_skel, "")));
856     _FDT((fdt_end_node(fdt_skel)));
857     _FDT((fdt_finish(fdt_skel)));
858     fdt = g_malloc0(size);
859     _FDT((fdt_open_into(fdt_skel, fdt, size)));
860     g_free(fdt_skel);
861 
862     /* Fixup cpu nodes */
863     if (cpu_update) {
864         _FDT((spapr_fixup_cpu_dt(fdt, spapr)));
865     }
866 
867     /* Generate memory nodes or ibm,dynamic-reconfiguration-memory node */
868     if (memory_update && smc->dr_lmb_enabled) {
869         _FDT((spapr_populate_drconf_memory(spapr, fdt)));
870     }
871 
872     /* Pack resulting tree */
873     _FDT((fdt_pack(fdt)));
874 
875     if (fdt_totalsize(fdt) + sizeof(hdr) > size) {
876         trace_spapr_cas_failed(size);
877         return -1;
878     }
879 
880     cpu_physical_memory_write(addr, &hdr, sizeof(hdr));
881     cpu_physical_memory_write(addr + sizeof(hdr), fdt, fdt_totalsize(fdt));
882     trace_spapr_cas_continue(fdt_totalsize(fdt) + sizeof(hdr));
883     g_free(fdt);
884 
885     return 0;
886 }
887 
888 static void spapr_finalize_fdt(sPAPRMachineState *spapr,
889                                hwaddr fdt_addr,
890                                hwaddr rtas_addr,
891                                hwaddr rtas_size)
892 {
893     MachineState *machine = MACHINE(qdev_get_machine());
894     sPAPRMachineClass *smc = SPAPR_MACHINE_GET_CLASS(machine);
895     const char *boot_device = machine->boot_order;
896     int ret, i;
897     size_t cb = 0;
898     char *bootlist;
899     void *fdt;
900     sPAPRPHBState *phb;
901 
902     fdt = g_malloc(FDT_MAX_SIZE);
903 
904     /* open out the base tree into a temp buffer for the final tweaks */
905     _FDT((fdt_open_into(spapr->fdt_skel, fdt, FDT_MAX_SIZE)));
906 
907     ret = spapr_populate_memory(spapr, fdt);
908     if (ret < 0) {
909         fprintf(stderr, "couldn't setup memory nodes in fdt\n");
910         exit(1);
911     }
912 
913     ret = spapr_populate_vdevice(spapr->vio_bus, fdt);
914     if (ret < 0) {
915         fprintf(stderr, "couldn't setup vio devices in fdt\n");
916         exit(1);
917     }
918 
919     if (object_resolve_path_type("", TYPE_SPAPR_RNG, NULL)) {
920         ret = spapr_rng_populate_dt(fdt);
921         if (ret < 0) {
922             fprintf(stderr, "could not set up rng device in the fdt\n");
923             exit(1);
924         }
925     }
926 
927     QLIST_FOREACH(phb, &spapr->phbs, list) {
928         ret = spapr_populate_pci_dt(phb, PHANDLE_XICP, fdt);
929     }
930 
931     if (ret < 0) {
932         fprintf(stderr, "couldn't setup PCI devices in fdt\n");
933         exit(1);
934     }
935 
936     /* RTAS */
937     ret = spapr_rtas_device_tree_setup(fdt, rtas_addr, rtas_size);
938     if (ret < 0) {
939         fprintf(stderr, "Couldn't set up RTAS device tree properties\n");
940     }
941 
942     /* cpus */
943     spapr_populate_cpus_dt_node(fdt, spapr);
944 
945     bootlist = get_boot_devices_list(&cb, true);
946     if (cb && bootlist) {
947         int offset = fdt_path_offset(fdt, "/chosen");
948         if (offset < 0) {
949             exit(1);
950         }
951         for (i = 0; i < cb; i++) {
952             if (bootlist[i] == '\n') {
953                 bootlist[i] = ' ';
954             }
955 
956         }
957         ret = fdt_setprop_string(fdt, offset, "qemu,boot-list", bootlist);
958     }
959 
960     if (boot_device && strlen(boot_device)) {
961         int offset = fdt_path_offset(fdt, "/chosen");
962 
963         if (offset < 0) {
964             exit(1);
965         }
966         fdt_setprop_string(fdt, offset, "qemu,boot-device", boot_device);
967     }
968 
969     if (!spapr->has_graphics) {
970         spapr_populate_chosen_stdout(fdt, spapr->vio_bus);
971     }
972 
973     if (smc->dr_lmb_enabled) {
974         _FDT(spapr_drc_populate_dt(fdt, 0, NULL, SPAPR_DR_CONNECTOR_TYPE_LMB));
975     }
976 
977     _FDT((fdt_pack(fdt)));
978 
979     if (fdt_totalsize(fdt) > FDT_MAX_SIZE) {
980         error_report("FDT too big ! 0x%x bytes (max is 0x%x)",
981                      fdt_totalsize(fdt), FDT_MAX_SIZE);
982         exit(1);
983     }
984 
985     qemu_fdt_dumpdtb(fdt, fdt_totalsize(fdt));
986     cpu_physical_memory_write(fdt_addr, fdt, fdt_totalsize(fdt));
987 
988     g_free(bootlist);
989     g_free(fdt);
990 }
991 
992 static uint64_t translate_kernel_address(void *opaque, uint64_t addr)
993 {
994     return (addr & 0x0fffffff) + KERNEL_LOAD_ADDR;
995 }
996 
997 static void emulate_spapr_hypercall(PowerPCCPU *cpu)
998 {
999     CPUPPCState *env = &cpu->env;
1000 
1001     if (msr_pr) {
1002         hcall_dprintf("Hypercall made with MSR[PR]=1\n");
1003         env->gpr[3] = H_PRIVILEGE;
1004     } else {
1005         env->gpr[3] = spapr_hypercall(cpu, env->gpr[3], &env->gpr[4]);
1006     }
1007 }
1008 
1009 #define HPTE(_table, _i)   (void *)(((uint64_t *)(_table)) + ((_i) * 2))
1010 #define HPTE_VALID(_hpte)  (tswap64(*((uint64_t *)(_hpte))) & HPTE64_V_VALID)
1011 #define HPTE_DIRTY(_hpte)  (tswap64(*((uint64_t *)(_hpte))) & HPTE64_V_HPTE_DIRTY)
1012 #define CLEAN_HPTE(_hpte)  ((*(uint64_t *)(_hpte)) &= tswap64(~HPTE64_V_HPTE_DIRTY))
1013 #define DIRTY_HPTE(_hpte)  ((*(uint64_t *)(_hpte)) |= tswap64(HPTE64_V_HPTE_DIRTY))
1014 
1015 static void spapr_alloc_htab(sPAPRMachineState *spapr)
1016 {
1017     long shift;
1018     int index;
1019 
1020     /* allocate hash page table.  For now we always make this 16mb,
1021      * later we should probably make it scale to the size of guest
1022      * RAM */
1023 
1024     shift = kvmppc_reset_htab(spapr->htab_shift);
1025     if (shift < 0) {
1026         /*
1027          * For HV KVM, host kernel will return -ENOMEM when requested
1028          * HTAB size can't be allocated.
1029          */
1030         error_setg(&error_abort, "Failed to allocate HTAB of requested size, try with smaller maxmem");
1031     } else if (shift > 0) {
1032         /*
1033          * Kernel handles htab, we don't need to allocate one
1034          *
1035          * Older kernels can fall back to lower HTAB shift values,
1036          * but we don't allow booting of such guests.
1037          */
1038         if (shift != spapr->htab_shift) {
1039             error_setg(&error_abort, "Failed to allocate HTAB of requested size, try with smaller maxmem");
1040         }
1041 
1042         spapr->htab_shift = shift;
1043         kvmppc_kern_htab = true;
1044     } else {
1045         /* Allocate htab */
1046         spapr->htab = qemu_memalign(HTAB_SIZE(spapr), HTAB_SIZE(spapr));
1047 
1048         /* And clear it */
1049         memset(spapr->htab, 0, HTAB_SIZE(spapr));
1050 
1051         for (index = 0; index < HTAB_SIZE(spapr) / HASH_PTE_SIZE_64; index++) {
1052             DIRTY_HPTE(HPTE(spapr->htab, index));
1053         }
1054     }
1055 }
1056 
1057 /*
1058  * Clear HTAB entries during reset.
1059  *
1060  * If host kernel has allocated HTAB, KVM_PPC_ALLOCATE_HTAB ioctl is
1061  * used to clear HTAB. Otherwise QEMU-allocated HTAB is cleared manually.
1062  */
1063 static void spapr_reset_htab(sPAPRMachineState *spapr)
1064 {
1065     long shift;
1066     int index;
1067 
1068     shift = kvmppc_reset_htab(spapr->htab_shift);
1069     if (shift < 0) {
1070         error_setg(&error_abort, "Failed to reset HTAB");
1071     } else if (shift > 0) {
1072         if (shift != spapr->htab_shift) {
1073             error_setg(&error_abort, "Requested HTAB allocation failed during reset");
1074         }
1075 
1076         /* Tell readers to update their file descriptor */
1077         if (spapr->htab_fd >= 0) {
1078             spapr->htab_fd_stale = true;
1079         }
1080     } else {
1081         memset(spapr->htab, 0, HTAB_SIZE(spapr));
1082 
1083         for (index = 0; index < HTAB_SIZE(spapr) / HASH_PTE_SIZE_64; index++) {
1084             DIRTY_HPTE(HPTE(spapr->htab, index));
1085         }
1086     }
1087 
1088     /* Update the RMA size if necessary */
1089     if (spapr->vrma_adjust) {
1090         spapr->rma_size = kvmppc_rma_size(spapr_node0_size(),
1091                                           spapr->htab_shift);
1092     }
1093 }
1094 
1095 static int find_unknown_sysbus_device(SysBusDevice *sbdev, void *opaque)
1096 {
1097     bool matched = false;
1098 
1099     if (object_dynamic_cast(OBJECT(sbdev), TYPE_SPAPR_PCI_HOST_BRIDGE)) {
1100         matched = true;
1101     }
1102 
1103     if (!matched) {
1104         error_report("Device %s is not supported by this machine yet.",
1105                      qdev_fw_name(DEVICE(sbdev)));
1106         exit(1);
1107     }
1108 
1109     return 0;
1110 }
1111 
1112 /*
1113  * A guest reset will cause spapr->htab_fd to become stale if being used.
1114  * Reopen the file descriptor to make sure the whole HTAB is properly read.
1115  */
1116 static int spapr_check_htab_fd(sPAPRMachineState *spapr)
1117 {
1118     int rc = 0;
1119 
1120     if (spapr->htab_fd_stale) {
1121         close(spapr->htab_fd);
1122         spapr->htab_fd = kvmppc_get_htab_fd(false);
1123         if (spapr->htab_fd < 0) {
1124             error_report("Unable to open fd for reading hash table from KVM: "
1125                          "%s", strerror(errno));
1126             rc = -1;
1127         }
1128         spapr->htab_fd_stale = false;
1129     }
1130 
1131     return rc;
1132 }
1133 
1134 static void ppc_spapr_reset(void)
1135 {
1136     sPAPRMachineState *spapr = SPAPR_MACHINE(qdev_get_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     /* Reset the hash table & recalc the RMA */
1144     spapr_reset_htab(spapr);
1145 
1146     qemu_devices_reset();
1147 
1148     /*
1149      * We place the device tree and RTAS just below either the top of the RMA,
1150      * or just below 2GB, whichever is lowere, so that it can be
1151      * processed with 32-bit real mode code if necessary
1152      */
1153     rtas_limit = MIN(spapr->rma_size, RTAS_MAX_ADDR);
1154     spapr->rtas_addr = rtas_limit - RTAS_MAX_SIZE;
1155     spapr->fdt_addr = spapr->rtas_addr - FDT_MAX_SIZE;
1156 
1157     /* Load the fdt */
1158     spapr_finalize_fdt(spapr, spapr->fdt_addr, spapr->rtas_addr,
1159                        spapr->rtas_size);
1160 
1161     /* Copy RTAS over */
1162     cpu_physical_memory_write(spapr->rtas_addr, spapr->rtas_blob,
1163                               spapr->rtas_size);
1164 
1165     /* Set up the entry state */
1166     first_ppc_cpu = POWERPC_CPU(first_cpu);
1167     first_ppc_cpu->env.gpr[3] = spapr->fdt_addr;
1168     first_ppc_cpu->env.gpr[5] = 0;
1169     first_cpu->halted = 0;
1170     first_ppc_cpu->env.nip = SPAPR_ENTRY_POINT;
1171 
1172 }
1173 
1174 static void spapr_cpu_reset(void *opaque)
1175 {
1176     sPAPRMachineState *spapr = SPAPR_MACHINE(qdev_get_machine());
1177     PowerPCCPU *cpu = opaque;
1178     CPUState *cs = CPU(cpu);
1179     CPUPPCState *env = &cpu->env;
1180 
1181     cpu_reset(cs);
1182 
1183     /* All CPUs start halted.  CPU0 is unhalted from the machine level
1184      * reset code and the rest are explicitly started up by the guest
1185      * using an RTAS call */
1186     cs->halted = 1;
1187 
1188     env->spr[SPR_HIOR] = 0;
1189 
1190     env->external_htab = (uint8_t *)spapr->htab;
1191     if (kvm_enabled() && !env->external_htab) {
1192         /*
1193          * HV KVM, set external_htab to 1 so our ppc_hash64_load_hpte*
1194          * functions do the right thing.
1195          */
1196         env->external_htab = (void *)1;
1197     }
1198     env->htab_base = -1;
1199     /*
1200      * htab_mask is the mask used to normalize hash value to PTEG index.
1201      * htab_shift is log2 of hash table size.
1202      * We have 8 hpte per group, and each hpte is 16 bytes.
1203      * ie have 128 bytes per hpte entry.
1204      */
1205     env->htab_mask = (1ULL << (spapr->htab_shift - 7)) - 1;
1206     env->spr[SPR_SDR1] = (target_ulong)(uintptr_t)spapr->htab |
1207         (spapr->htab_shift - 18);
1208 }
1209 
1210 static void spapr_create_nvram(sPAPRMachineState *spapr)
1211 {
1212     DeviceState *dev = qdev_create(&spapr->vio_bus->bus, "spapr-nvram");
1213     DriveInfo *dinfo = drive_get(IF_PFLASH, 0, 0);
1214 
1215     if (dinfo) {
1216         qdev_prop_set_drive_nofail(dev, "drive", blk_by_legacy_dinfo(dinfo));
1217     }
1218 
1219     qdev_init_nofail(dev);
1220 
1221     spapr->nvram = (struct sPAPRNVRAM *)dev;
1222 }
1223 
1224 static void spapr_rtc_create(sPAPRMachineState *spapr)
1225 {
1226     DeviceState *dev = qdev_create(NULL, TYPE_SPAPR_RTC);
1227 
1228     qdev_init_nofail(dev);
1229     spapr->rtc = dev;
1230 
1231     object_property_add_alias(qdev_get_machine(), "rtc-time",
1232                               OBJECT(spapr->rtc), "date", NULL);
1233 }
1234 
1235 /* Returns whether we want to use VGA or not */
1236 static int spapr_vga_init(PCIBus *pci_bus)
1237 {
1238     switch (vga_interface_type) {
1239     case VGA_NONE:
1240         return false;
1241     case VGA_DEVICE:
1242         return true;
1243     case VGA_STD:
1244     case VGA_VIRTIO:
1245         return pci_vga_init(pci_bus) != NULL;
1246     default:
1247         fprintf(stderr, "This vga model is not supported,"
1248                 "currently it only supports -vga std\n");
1249         exit(0);
1250     }
1251 }
1252 
1253 static int spapr_post_load(void *opaque, int version_id)
1254 {
1255     sPAPRMachineState *spapr = (sPAPRMachineState *)opaque;
1256     int err = 0;
1257 
1258     /* In earlier versions, there was no separate qdev for the PAPR
1259      * RTC, so the RTC offset was stored directly in sPAPREnvironment.
1260      * So when migrating from those versions, poke the incoming offset
1261      * value into the RTC device */
1262     if (version_id < 3) {
1263         err = spapr_rtc_import_offset(spapr->rtc, spapr->rtc_offset);
1264     }
1265 
1266     return err;
1267 }
1268 
1269 static bool version_before_3(void *opaque, int version_id)
1270 {
1271     return version_id < 3;
1272 }
1273 
1274 static const VMStateDescription vmstate_spapr = {
1275     .name = "spapr",
1276     .version_id = 3,
1277     .minimum_version_id = 1,
1278     .post_load = spapr_post_load,
1279     .fields = (VMStateField[]) {
1280         /* used to be @next_irq */
1281         VMSTATE_UNUSED_BUFFER(version_before_3, 0, 4),
1282 
1283         /* RTC offset */
1284         VMSTATE_UINT64_TEST(rtc_offset, sPAPRMachineState, version_before_3),
1285 
1286         VMSTATE_PPC_TIMEBASE_V(tb, sPAPRMachineState, 2),
1287         VMSTATE_END_OF_LIST()
1288     },
1289 };
1290 
1291 static int htab_save_setup(QEMUFile *f, void *opaque)
1292 {
1293     sPAPRMachineState *spapr = opaque;
1294 
1295     /* "Iteration" header */
1296     qemu_put_be32(f, spapr->htab_shift);
1297 
1298     if (spapr->htab) {
1299         spapr->htab_save_index = 0;
1300         spapr->htab_first_pass = true;
1301     } else {
1302         assert(kvm_enabled());
1303 
1304         spapr->htab_fd = kvmppc_get_htab_fd(false);
1305         spapr->htab_fd_stale = false;
1306         if (spapr->htab_fd < 0) {
1307             fprintf(stderr, "Unable to open fd for reading hash table from KVM: %s\n",
1308                     strerror(errno));
1309             return -1;
1310         }
1311     }
1312 
1313 
1314     return 0;
1315 }
1316 
1317 static void htab_save_first_pass(QEMUFile *f, sPAPRMachineState *spapr,
1318                                  int64_t max_ns)
1319 {
1320     int htabslots = HTAB_SIZE(spapr) / HASH_PTE_SIZE_64;
1321     int index = spapr->htab_save_index;
1322     int64_t starttime = qemu_clock_get_ns(QEMU_CLOCK_REALTIME);
1323 
1324     assert(spapr->htab_first_pass);
1325 
1326     do {
1327         int chunkstart;
1328 
1329         /* Consume invalid HPTEs */
1330         while ((index < htabslots)
1331                && !HPTE_VALID(HPTE(spapr->htab, index))) {
1332             index++;
1333             CLEAN_HPTE(HPTE(spapr->htab, index));
1334         }
1335 
1336         /* Consume valid HPTEs */
1337         chunkstart = index;
1338         while ((index < htabslots) && (index - chunkstart < USHRT_MAX)
1339                && HPTE_VALID(HPTE(spapr->htab, index))) {
1340             index++;
1341             CLEAN_HPTE(HPTE(spapr->htab, index));
1342         }
1343 
1344         if (index > chunkstart) {
1345             int n_valid = index - chunkstart;
1346 
1347             qemu_put_be32(f, chunkstart);
1348             qemu_put_be16(f, n_valid);
1349             qemu_put_be16(f, 0);
1350             qemu_put_buffer(f, HPTE(spapr->htab, chunkstart),
1351                             HASH_PTE_SIZE_64 * n_valid);
1352 
1353             if ((qemu_clock_get_ns(QEMU_CLOCK_REALTIME) - starttime) > max_ns) {
1354                 break;
1355             }
1356         }
1357     } while ((index < htabslots) && !qemu_file_rate_limit(f));
1358 
1359     if (index >= htabslots) {
1360         assert(index == htabslots);
1361         index = 0;
1362         spapr->htab_first_pass = false;
1363     }
1364     spapr->htab_save_index = index;
1365 }
1366 
1367 static int htab_save_later_pass(QEMUFile *f, sPAPRMachineState *spapr,
1368                                 int64_t max_ns)
1369 {
1370     bool final = max_ns < 0;
1371     int htabslots = HTAB_SIZE(spapr) / HASH_PTE_SIZE_64;
1372     int examined = 0, sent = 0;
1373     int index = spapr->htab_save_index;
1374     int64_t starttime = qemu_clock_get_ns(QEMU_CLOCK_REALTIME);
1375 
1376     assert(!spapr->htab_first_pass);
1377 
1378     do {
1379         int chunkstart, invalidstart;
1380 
1381         /* Consume non-dirty HPTEs */
1382         while ((index < htabslots)
1383                && !HPTE_DIRTY(HPTE(spapr->htab, index))) {
1384             index++;
1385             examined++;
1386         }
1387 
1388         chunkstart = index;
1389         /* Consume valid dirty HPTEs */
1390         while ((index < htabslots) && (index - chunkstart < USHRT_MAX)
1391                && HPTE_DIRTY(HPTE(spapr->htab, index))
1392                && HPTE_VALID(HPTE(spapr->htab, index))) {
1393             CLEAN_HPTE(HPTE(spapr->htab, index));
1394             index++;
1395             examined++;
1396         }
1397 
1398         invalidstart = index;
1399         /* Consume invalid dirty HPTEs */
1400         while ((index < htabslots) && (index - invalidstart < USHRT_MAX)
1401                && HPTE_DIRTY(HPTE(spapr->htab, index))
1402                && !HPTE_VALID(HPTE(spapr->htab, index))) {
1403             CLEAN_HPTE(HPTE(spapr->htab, index));
1404             index++;
1405             examined++;
1406         }
1407 
1408         if (index > chunkstart) {
1409             int n_valid = invalidstart - chunkstart;
1410             int n_invalid = index - invalidstart;
1411 
1412             qemu_put_be32(f, chunkstart);
1413             qemu_put_be16(f, n_valid);
1414             qemu_put_be16(f, n_invalid);
1415             qemu_put_buffer(f, HPTE(spapr->htab, chunkstart),
1416                             HASH_PTE_SIZE_64 * n_valid);
1417             sent += index - chunkstart;
1418 
1419             if (!final && (qemu_clock_get_ns(QEMU_CLOCK_REALTIME) - starttime) > max_ns) {
1420                 break;
1421             }
1422         }
1423 
1424         if (examined >= htabslots) {
1425             break;
1426         }
1427 
1428         if (index >= htabslots) {
1429             assert(index == htabslots);
1430             index = 0;
1431         }
1432     } while ((examined < htabslots) && (!qemu_file_rate_limit(f) || final));
1433 
1434     if (index >= htabslots) {
1435         assert(index == htabslots);
1436         index = 0;
1437     }
1438 
1439     spapr->htab_save_index = index;
1440 
1441     return (examined >= htabslots) && (sent == 0) ? 1 : 0;
1442 }
1443 
1444 #define MAX_ITERATION_NS    5000000 /* 5 ms */
1445 #define MAX_KVM_BUF_SIZE    2048
1446 
1447 static int htab_save_iterate(QEMUFile *f, void *opaque)
1448 {
1449     sPAPRMachineState *spapr = opaque;
1450     int rc = 0;
1451 
1452     /* Iteration header */
1453     qemu_put_be32(f, 0);
1454 
1455     if (!spapr->htab) {
1456         assert(kvm_enabled());
1457 
1458         rc = spapr_check_htab_fd(spapr);
1459         if (rc < 0) {
1460             return rc;
1461         }
1462 
1463         rc = kvmppc_save_htab(f, spapr->htab_fd,
1464                               MAX_KVM_BUF_SIZE, MAX_ITERATION_NS);
1465         if (rc < 0) {
1466             return rc;
1467         }
1468     } else  if (spapr->htab_first_pass) {
1469         htab_save_first_pass(f, spapr, MAX_ITERATION_NS);
1470     } else {
1471         rc = htab_save_later_pass(f, spapr, MAX_ITERATION_NS);
1472     }
1473 
1474     /* End marker */
1475     qemu_put_be32(f, 0);
1476     qemu_put_be16(f, 0);
1477     qemu_put_be16(f, 0);
1478 
1479     return rc;
1480 }
1481 
1482 static int htab_save_complete(QEMUFile *f, void *opaque)
1483 {
1484     sPAPRMachineState *spapr = opaque;
1485 
1486     /* Iteration header */
1487     qemu_put_be32(f, 0);
1488 
1489     if (!spapr->htab) {
1490         int rc;
1491 
1492         assert(kvm_enabled());
1493 
1494         rc = spapr_check_htab_fd(spapr);
1495         if (rc < 0) {
1496             return rc;
1497         }
1498 
1499         rc = kvmppc_save_htab(f, spapr->htab_fd, MAX_KVM_BUF_SIZE, -1);
1500         if (rc < 0) {
1501             return rc;
1502         }
1503         close(spapr->htab_fd);
1504         spapr->htab_fd = -1;
1505     } else {
1506         htab_save_later_pass(f, spapr, -1);
1507     }
1508 
1509     /* End marker */
1510     qemu_put_be32(f, 0);
1511     qemu_put_be16(f, 0);
1512     qemu_put_be16(f, 0);
1513 
1514     return 0;
1515 }
1516 
1517 static int htab_load(QEMUFile *f, void *opaque, int version_id)
1518 {
1519     sPAPRMachineState *spapr = opaque;
1520     uint32_t section_hdr;
1521     int fd = -1;
1522 
1523     if (version_id < 1 || version_id > 1) {
1524         fprintf(stderr, "htab_load() bad version\n");
1525         return -EINVAL;
1526     }
1527 
1528     section_hdr = qemu_get_be32(f);
1529 
1530     if (section_hdr) {
1531         /* First section, just the hash shift */
1532         if (spapr->htab_shift != section_hdr) {
1533             error_report("htab_shift mismatch: source %d target %d",
1534                          section_hdr, spapr->htab_shift);
1535             return -EINVAL;
1536         }
1537         return 0;
1538     }
1539 
1540     if (!spapr->htab) {
1541         assert(kvm_enabled());
1542 
1543         fd = kvmppc_get_htab_fd(true);
1544         if (fd < 0) {
1545             fprintf(stderr, "Unable to open fd to restore KVM hash table: %s\n",
1546                     strerror(errno));
1547         }
1548     }
1549 
1550     while (true) {
1551         uint32_t index;
1552         uint16_t n_valid, n_invalid;
1553 
1554         index = qemu_get_be32(f);
1555         n_valid = qemu_get_be16(f);
1556         n_invalid = qemu_get_be16(f);
1557 
1558         if ((index == 0) && (n_valid == 0) && (n_invalid == 0)) {
1559             /* End of Stream */
1560             break;
1561         }
1562 
1563         if ((index + n_valid + n_invalid) >
1564             (HTAB_SIZE(spapr) / HASH_PTE_SIZE_64)) {
1565             /* Bad index in stream */
1566             fprintf(stderr, "htab_load() bad index %d (%hd+%hd entries) "
1567                     "in htab stream (htab_shift=%d)\n", index, n_valid, n_invalid,
1568                     spapr->htab_shift);
1569             return -EINVAL;
1570         }
1571 
1572         if (spapr->htab) {
1573             if (n_valid) {
1574                 qemu_get_buffer(f, HPTE(spapr->htab, index),
1575                                 HASH_PTE_SIZE_64 * n_valid);
1576             }
1577             if (n_invalid) {
1578                 memset(HPTE(spapr->htab, index + n_valid), 0,
1579                        HASH_PTE_SIZE_64 * n_invalid);
1580             }
1581         } else {
1582             int rc;
1583 
1584             assert(fd >= 0);
1585 
1586             rc = kvmppc_load_htab_chunk(f, fd, index, n_valid, n_invalid);
1587             if (rc < 0) {
1588                 return rc;
1589             }
1590         }
1591     }
1592 
1593     if (!spapr->htab) {
1594         assert(fd >= 0);
1595         close(fd);
1596     }
1597 
1598     return 0;
1599 }
1600 
1601 static SaveVMHandlers savevm_htab_handlers = {
1602     .save_live_setup = htab_save_setup,
1603     .save_live_iterate = htab_save_iterate,
1604     .save_live_complete_precopy = htab_save_complete,
1605     .load_state = htab_load,
1606 };
1607 
1608 static void spapr_boot_set(void *opaque, const char *boot_device,
1609                            Error **errp)
1610 {
1611     MachineState *machine = MACHINE(qdev_get_machine());
1612     machine->boot_order = g_strdup(boot_device);
1613 }
1614 
1615 static void spapr_cpu_init(sPAPRMachineState *spapr, PowerPCCPU *cpu)
1616 {
1617     CPUPPCState *env = &cpu->env;
1618 
1619     /* Set time-base frequency to 512 MHz */
1620     cpu_ppc_tb_init(env, TIMEBASE_FREQ);
1621 
1622     /* PAPR always has exception vectors in RAM not ROM. To ensure this,
1623      * MSR[IP] should never be set.
1624      */
1625     env->msr_mask &= ~(1 << 6);
1626 
1627     /* Tell KVM that we're in PAPR mode */
1628     if (kvm_enabled()) {
1629         kvmppc_set_papr(cpu);
1630     }
1631 
1632     if (cpu->max_compat) {
1633         if (ppc_set_compat(cpu, cpu->max_compat) < 0) {
1634             exit(1);
1635         }
1636     }
1637 
1638     xics_cpu_setup(spapr->icp, cpu);
1639 
1640     qemu_register_reset(spapr_cpu_reset, cpu);
1641 }
1642 
1643 /*
1644  * Reset routine for LMB DR devices.
1645  *
1646  * Unlike PCI DR devices, LMB DR devices explicitly register this reset
1647  * routine. Reset for PCI DR devices will be handled by PHB reset routine
1648  * when it walks all its children devices. LMB devices reset occurs
1649  * as part of spapr_ppc_reset().
1650  */
1651 static void spapr_drc_reset(void *opaque)
1652 {
1653     sPAPRDRConnector *drc = opaque;
1654     DeviceState *d = DEVICE(drc);
1655 
1656     if (d) {
1657         device_reset(d);
1658     }
1659 }
1660 
1661 static void spapr_create_lmb_dr_connectors(sPAPRMachineState *spapr)
1662 {
1663     MachineState *machine = MACHINE(spapr);
1664     uint64_t lmb_size = SPAPR_MEMORY_BLOCK_SIZE;
1665     uint32_t nr_lmbs = (machine->maxram_size - machine->ram_size)/lmb_size;
1666     int i;
1667 
1668     for (i = 0; i < nr_lmbs; i++) {
1669         sPAPRDRConnector *drc;
1670         uint64_t addr;
1671 
1672         addr = i * lmb_size + spapr->hotplug_memory.base;
1673         drc = spapr_dr_connector_new(OBJECT(spapr), SPAPR_DR_CONNECTOR_TYPE_LMB,
1674                                      addr/lmb_size);
1675         qemu_register_reset(spapr_drc_reset, drc);
1676     }
1677 }
1678 
1679 /*
1680  * If RAM size, maxmem size and individual node mem sizes aren't aligned
1681  * to SPAPR_MEMORY_BLOCK_SIZE(256MB), then refuse to start the guest
1682  * since we can't support such unaligned sizes with DRCONF_MEMORY.
1683  */
1684 static void spapr_validate_node_memory(MachineState *machine)
1685 {
1686     int i;
1687 
1688     if (machine->maxram_size % SPAPR_MEMORY_BLOCK_SIZE ||
1689         machine->ram_size % SPAPR_MEMORY_BLOCK_SIZE) {
1690         error_report("Can't support memory configuration where RAM size "
1691                      "0x" RAM_ADDR_FMT " or maxmem size "
1692                      "0x" RAM_ADDR_FMT " isn't aligned to %llu MB",
1693                      machine->ram_size, machine->maxram_size,
1694                      SPAPR_MEMORY_BLOCK_SIZE/M_BYTE);
1695         exit(EXIT_FAILURE);
1696     }
1697 
1698     for (i = 0; i < nb_numa_nodes; i++) {
1699         if (numa_info[i].node_mem % SPAPR_MEMORY_BLOCK_SIZE) {
1700             error_report("Can't support memory configuration where memory size"
1701                          " %" PRIx64 " of node %d isn't aligned to %llu MB",
1702                          numa_info[i].node_mem, i,
1703                          SPAPR_MEMORY_BLOCK_SIZE/M_BYTE);
1704             exit(EXIT_FAILURE);
1705         }
1706     }
1707 }
1708 
1709 /* pSeries LPAR / sPAPR hardware init */
1710 static void ppc_spapr_init(MachineState *machine)
1711 {
1712     sPAPRMachineState *spapr = SPAPR_MACHINE(machine);
1713     sPAPRMachineClass *smc = SPAPR_MACHINE_GET_CLASS(machine);
1714     const char *kernel_filename = machine->kernel_filename;
1715     const char *kernel_cmdline = machine->kernel_cmdline;
1716     const char *initrd_filename = machine->initrd_filename;
1717     PowerPCCPU *cpu;
1718     PCIHostState *phb;
1719     int i;
1720     MemoryRegion *sysmem = get_system_memory();
1721     MemoryRegion *ram = g_new(MemoryRegion, 1);
1722     MemoryRegion *rma_region;
1723     void *rma = NULL;
1724     hwaddr rma_alloc_size;
1725     hwaddr node0_size = spapr_node0_size();
1726     uint32_t initrd_base = 0;
1727     long kernel_size = 0, initrd_size = 0;
1728     long load_limit, fw_size;
1729     bool kernel_le = false;
1730     char *filename;
1731 
1732     msi_supported = true;
1733 
1734     QLIST_INIT(&spapr->phbs);
1735 
1736     cpu_ppc_hypercall = emulate_spapr_hypercall;
1737 
1738     /* Allocate RMA if necessary */
1739     rma_alloc_size = kvmppc_alloc_rma(&rma);
1740 
1741     if (rma_alloc_size == -1) {
1742         error_report("Unable to create RMA");
1743         exit(1);
1744     }
1745 
1746     if (rma_alloc_size && (rma_alloc_size < node0_size)) {
1747         spapr->rma_size = rma_alloc_size;
1748     } else {
1749         spapr->rma_size = node0_size;
1750 
1751         /* With KVM, we don't actually know whether KVM supports an
1752          * unbounded RMA (PR KVM) or is limited by the hash table size
1753          * (HV KVM using VRMA), so we always assume the latter
1754          *
1755          * In that case, we also limit the initial allocations for RTAS
1756          * etc... to 256M since we have no way to know what the VRMA size
1757          * is going to be as it depends on the size of the hash table
1758          * isn't determined yet.
1759          */
1760         if (kvm_enabled()) {
1761             spapr->vrma_adjust = 1;
1762             spapr->rma_size = MIN(spapr->rma_size, 0x10000000);
1763         }
1764     }
1765 
1766     if (spapr->rma_size > node0_size) {
1767         fprintf(stderr, "Error: Numa node 0 has to span the RMA (%#08"HWADDR_PRIx")\n",
1768                 spapr->rma_size);
1769         exit(1);
1770     }
1771 
1772     /* Setup a load limit for the ramdisk leaving room for SLOF and FDT */
1773     load_limit = MIN(spapr->rma_size, RTAS_MAX_ADDR) - FW_OVERHEAD;
1774 
1775     /* We aim for a hash table of size 1/128 the size of RAM.  The
1776      * normal rule of thumb is 1/64 the size of RAM, but that's much
1777      * more than needed for the Linux guests we support. */
1778     spapr->htab_shift = 18; /* Minimum architected size */
1779     while (spapr->htab_shift <= 46) {
1780         if ((1ULL << (spapr->htab_shift + 7)) >= machine->maxram_size) {
1781             break;
1782         }
1783         spapr->htab_shift++;
1784     }
1785     spapr_alloc_htab(spapr);
1786 
1787     /* Set up Interrupt Controller before we create the VCPUs */
1788     spapr->icp = xics_system_init(machine,
1789                                   DIV_ROUND_UP(max_cpus * kvmppc_smt_threads(),
1790                                                smp_threads),
1791                                   XICS_IRQS);
1792 
1793     if (smc->dr_lmb_enabled) {
1794         spapr_validate_node_memory(machine);
1795     }
1796 
1797     /* init CPUs */
1798     if (machine->cpu_model == NULL) {
1799         machine->cpu_model = kvm_enabled() ? "host" : "POWER7";
1800     }
1801     for (i = 0; i < smp_cpus; i++) {
1802         cpu = cpu_ppc_init(machine->cpu_model);
1803         if (cpu == NULL) {
1804             fprintf(stderr, "Unable to find PowerPC CPU definition\n");
1805             exit(1);
1806         }
1807         spapr_cpu_init(spapr, cpu);
1808     }
1809 
1810     if (kvm_enabled()) {
1811         /* Enable H_LOGICAL_CI_* so SLOF can talk to in-kernel devices */
1812         kvmppc_enable_logical_ci_hcalls();
1813         kvmppc_enable_set_mode_hcall();
1814     }
1815 
1816     /* allocate RAM */
1817     memory_region_allocate_system_memory(ram, NULL, "ppc_spapr.ram",
1818                                          machine->ram_size);
1819     memory_region_add_subregion(sysmem, 0, ram);
1820 
1821     if (rma_alloc_size && rma) {
1822         rma_region = g_new(MemoryRegion, 1);
1823         memory_region_init_ram_ptr(rma_region, NULL, "ppc_spapr.rma",
1824                                    rma_alloc_size, rma);
1825         vmstate_register_ram_global(rma_region);
1826         memory_region_add_subregion(sysmem, 0, rma_region);
1827     }
1828 
1829     /* initialize hotplug memory address space */
1830     if (machine->ram_size < machine->maxram_size) {
1831         ram_addr_t hotplug_mem_size = machine->maxram_size - machine->ram_size;
1832 
1833         if (machine->ram_slots > SPAPR_MAX_RAM_SLOTS) {
1834             error_report("Specified number of memory slots %"PRIu64" exceeds max supported %d\n",
1835                          machine->ram_slots, SPAPR_MAX_RAM_SLOTS);
1836             exit(EXIT_FAILURE);
1837         }
1838 
1839         spapr->hotplug_memory.base = ROUND_UP(machine->ram_size,
1840                                               SPAPR_HOTPLUG_MEM_ALIGN);
1841         memory_region_init(&spapr->hotplug_memory.mr, OBJECT(spapr),
1842                            "hotplug-memory", hotplug_mem_size);
1843         memory_region_add_subregion(sysmem, spapr->hotplug_memory.base,
1844                                     &spapr->hotplug_memory.mr);
1845     }
1846 
1847     if (smc->dr_lmb_enabled) {
1848         spapr_create_lmb_dr_connectors(spapr);
1849     }
1850 
1851     filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, "spapr-rtas.bin");
1852     if (!filename) {
1853         error_report("Could not find LPAR rtas '%s'", "spapr-rtas.bin");
1854         exit(1);
1855     }
1856     spapr->rtas_size = get_image_size(filename);
1857     spapr->rtas_blob = g_malloc(spapr->rtas_size);
1858     if (load_image_size(filename, spapr->rtas_blob, spapr->rtas_size) < 0) {
1859         error_report("Could not load LPAR rtas '%s'", filename);
1860         exit(1);
1861     }
1862     if (spapr->rtas_size > RTAS_MAX_SIZE) {
1863         error_report("RTAS too big ! 0x%zx bytes (max is 0x%x)",
1864                      (size_t)spapr->rtas_size, RTAS_MAX_SIZE);
1865         exit(1);
1866     }
1867     g_free(filename);
1868 
1869     /* Set up EPOW events infrastructure */
1870     spapr_events_init(spapr);
1871 
1872     /* Set up the RTC RTAS interfaces */
1873     spapr_rtc_create(spapr);
1874 
1875     /* Set up VIO bus */
1876     spapr->vio_bus = spapr_vio_bus_init();
1877 
1878     for (i = 0; i < MAX_SERIAL_PORTS; i++) {
1879         if (serial_hds[i]) {
1880             spapr_vty_create(spapr->vio_bus, serial_hds[i]);
1881         }
1882     }
1883 
1884     /* We always have at least the nvram device on VIO */
1885     spapr_create_nvram(spapr);
1886 
1887     /* Set up PCI */
1888     spapr_pci_rtas_init();
1889 
1890     phb = spapr_create_phb(spapr, 0);
1891 
1892     for (i = 0; i < nb_nics; i++) {
1893         NICInfo *nd = &nd_table[i];
1894 
1895         if (!nd->model) {
1896             nd->model = g_strdup("ibmveth");
1897         }
1898 
1899         if (strcmp(nd->model, "ibmveth") == 0) {
1900             spapr_vlan_create(spapr->vio_bus, nd);
1901         } else {
1902             pci_nic_init_nofail(&nd_table[i], phb->bus, nd->model, NULL);
1903         }
1904     }
1905 
1906     for (i = 0; i <= drive_get_max_bus(IF_SCSI); i++) {
1907         spapr_vscsi_create(spapr->vio_bus);
1908     }
1909 
1910     /* Graphics */
1911     if (spapr_vga_init(phb->bus)) {
1912         spapr->has_graphics = true;
1913         machine->usb |= defaults_enabled() && !machine->usb_disabled;
1914     }
1915 
1916     if (machine->usb) {
1917         pci_create_simple(phb->bus, -1, "pci-ohci");
1918 
1919         if (spapr->has_graphics) {
1920             USBBus *usb_bus = usb_bus_find(-1);
1921 
1922             usb_create_simple(usb_bus, "usb-kbd");
1923             usb_create_simple(usb_bus, "usb-mouse");
1924         }
1925     }
1926 
1927     if (spapr->rma_size < (MIN_RMA_SLOF << 20)) {
1928         fprintf(stderr, "qemu: pSeries SLOF firmware requires >= "
1929                 "%ldM guest RMA (Real Mode Area memory)\n", MIN_RMA_SLOF);
1930         exit(1);
1931     }
1932 
1933     if (kernel_filename) {
1934         uint64_t lowaddr = 0;
1935 
1936         kernel_size = load_elf(kernel_filename, translate_kernel_address, NULL,
1937                                NULL, &lowaddr, NULL, 1, PPC_ELF_MACHINE, 0);
1938         if (kernel_size == ELF_LOAD_WRONG_ENDIAN) {
1939             kernel_size = load_elf(kernel_filename,
1940                                    translate_kernel_address, NULL,
1941                                    NULL, &lowaddr, NULL, 0, PPC_ELF_MACHINE, 0);
1942             kernel_le = kernel_size > 0;
1943         }
1944         if (kernel_size < 0) {
1945             fprintf(stderr, "qemu: error loading %s: %s\n",
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                 fprintf(stderr, "qemu: could not load initial ram disk '%s'\n",
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     object_property_add_str(obj, "kvm-type",
2098                             spapr_get_kvm_type, spapr_set_kvm_type, NULL);
2099     object_property_set_description(obj, "kvm-type",
2100                                     "Specifies the KVM virtualization mode (HV, PR)",
2101                                     NULL);
2102 }
2103 
2104 static void ppc_cpu_do_nmi_on_cpu(void *arg)
2105 {
2106     CPUState *cs = arg;
2107 
2108     cpu_synchronize_state(cs);
2109     ppc_cpu_do_system_reset(cs);
2110 }
2111 
2112 static void spapr_nmi(NMIState *n, int cpu_index, Error **errp)
2113 {
2114     CPUState *cs;
2115 
2116     CPU_FOREACH(cs) {
2117         async_run_on_cpu(cs, ppc_cpu_do_nmi_on_cpu, cs);
2118     }
2119 }
2120 
2121 static void spapr_add_lmbs(DeviceState *dev, uint64_t addr, uint64_t size,
2122                            uint32_t node, Error **errp)
2123 {
2124     sPAPRDRConnector *drc;
2125     sPAPRDRConnectorClass *drck;
2126     uint32_t nr_lmbs = size/SPAPR_MEMORY_BLOCK_SIZE;
2127     int i, fdt_offset, fdt_size;
2128     void *fdt;
2129 
2130     /*
2131      * Check for DRC connectors and send hotplug notification to the
2132      * guest only in case of hotplugged memory. This allows cold plugged
2133      * memory to be specified at boot time.
2134      */
2135     if (!dev->hotplugged) {
2136         return;
2137     }
2138 
2139     for (i = 0; i < nr_lmbs; i++) {
2140         drc = spapr_dr_connector_by_id(SPAPR_DR_CONNECTOR_TYPE_LMB,
2141                 addr/SPAPR_MEMORY_BLOCK_SIZE);
2142         g_assert(drc);
2143 
2144         fdt = create_device_tree(&fdt_size);
2145         fdt_offset = spapr_populate_memory_node(fdt, node, addr,
2146                                                 SPAPR_MEMORY_BLOCK_SIZE);
2147 
2148         drck = SPAPR_DR_CONNECTOR_GET_CLASS(drc);
2149         drck->attach(drc, dev, fdt, fdt_offset, !dev->hotplugged, errp);
2150         addr += SPAPR_MEMORY_BLOCK_SIZE;
2151     }
2152     spapr_hotplug_req_add_by_count(SPAPR_DR_CONNECTOR_TYPE_LMB, nr_lmbs);
2153 }
2154 
2155 static void spapr_memory_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
2156                               uint32_t node, Error **errp)
2157 {
2158     Error *local_err = NULL;
2159     sPAPRMachineState *ms = SPAPR_MACHINE(hotplug_dev);
2160     PCDIMMDevice *dimm = PC_DIMM(dev);
2161     PCDIMMDeviceClass *ddc = PC_DIMM_GET_CLASS(dimm);
2162     MemoryRegion *mr = ddc->get_memory_region(dimm);
2163     uint64_t align = memory_region_get_alignment(mr);
2164     uint64_t size = memory_region_size(mr);
2165     uint64_t addr;
2166 
2167     if (size % SPAPR_MEMORY_BLOCK_SIZE) {
2168         error_setg(&local_err, "Hotplugged memory size must be a multiple of "
2169                       "%lld MB", SPAPR_MEMORY_BLOCK_SIZE/M_BYTE);
2170         goto out;
2171     }
2172 
2173     pc_dimm_memory_plug(dev, &ms->hotplug_memory, mr, align, &local_err);
2174     if (local_err) {
2175         goto out;
2176     }
2177 
2178     addr = object_property_get_int(OBJECT(dimm), PC_DIMM_ADDR_PROP, &local_err);
2179     if (local_err) {
2180         pc_dimm_memory_unplug(dev, &ms->hotplug_memory, mr);
2181         goto out;
2182     }
2183 
2184     spapr_add_lmbs(dev, addr, size, node, &error_abort);
2185 
2186 out:
2187     error_propagate(errp, local_err);
2188 }
2189 
2190 static void spapr_machine_device_plug(HotplugHandler *hotplug_dev,
2191                                       DeviceState *dev, Error **errp)
2192 {
2193     sPAPRMachineClass *smc = SPAPR_MACHINE_GET_CLASS(qdev_get_machine());
2194 
2195     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2196         int node;
2197 
2198         if (!smc->dr_lmb_enabled) {
2199             error_setg(errp, "Memory hotplug not supported for this machine");
2200             return;
2201         }
2202         node = object_property_get_int(OBJECT(dev), PC_DIMM_NODE_PROP, errp);
2203         if (*errp) {
2204             return;
2205         }
2206 
2207         /*
2208          * Currently PowerPC kernel doesn't allow hot-adding memory to
2209          * memory-less node, but instead will silently add the memory
2210          * to the first node that has some memory. This causes two
2211          * unexpected behaviours for the user.
2212          *
2213          * - Memory gets hotplugged to a different node than what the user
2214          *   specified.
2215          * - Since pc-dimm subsystem in QEMU still thinks that memory belongs
2216          *   to memory-less node, a reboot will set things accordingly
2217          *   and the previously hotplugged memory now ends in the right node.
2218          *   This appears as if some memory moved from one node to another.
2219          *
2220          * So until kernel starts supporting memory hotplug to memory-less
2221          * nodes, just prevent such attempts upfront in QEMU.
2222          */
2223         if (nb_numa_nodes && !numa_info[node].node_mem) {
2224             error_setg(errp, "Can't hotplug memory to memory-less node %d",
2225                        node);
2226             return;
2227         }
2228 
2229         spapr_memory_plug(hotplug_dev, dev, node, errp);
2230     }
2231 }
2232 
2233 static void spapr_machine_device_unplug(HotplugHandler *hotplug_dev,
2234                                       DeviceState *dev, Error **errp)
2235 {
2236     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2237         error_setg(errp, "Memory hot unplug not supported by sPAPR");
2238     }
2239 }
2240 
2241 static HotplugHandler *spapr_get_hotpug_handler(MachineState *machine,
2242                                              DeviceState *dev)
2243 {
2244     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2245         return HOTPLUG_HANDLER(machine);
2246     }
2247     return NULL;
2248 }
2249 
2250 static unsigned spapr_cpu_index_to_socket_id(unsigned cpu_index)
2251 {
2252     /* Allocate to NUMA nodes on a "socket" basis (not that concept of
2253      * socket means much for the paravirtualized PAPR platform) */
2254     return cpu_index / smp_threads / smp_cores;
2255 }
2256 
2257 static void spapr_machine_class_init(ObjectClass *oc, void *data)
2258 {
2259     MachineClass *mc = MACHINE_CLASS(oc);
2260     sPAPRMachineClass *smc = SPAPR_MACHINE_CLASS(oc);
2261     FWPathProviderClass *fwc = FW_PATH_PROVIDER_CLASS(oc);
2262     NMIClass *nc = NMI_CLASS(oc);
2263     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
2264 
2265     mc->init = ppc_spapr_init;
2266     mc->reset = ppc_spapr_reset;
2267     mc->block_default_type = IF_SCSI;
2268     mc->max_cpus = MAX_CPUMASK_BITS;
2269     mc->no_parallel = 1;
2270     mc->default_boot_order = "";
2271     mc->default_ram_size = 512 * M_BYTE;
2272     mc->kvm_type = spapr_kvm_type;
2273     mc->has_dynamic_sysbus = true;
2274     mc->pci_allow_0_address = true;
2275     mc->get_hotplug_handler = spapr_get_hotpug_handler;
2276     hc->plug = spapr_machine_device_plug;
2277     hc->unplug = spapr_machine_device_unplug;
2278     mc->cpu_index_to_socket_id = spapr_cpu_index_to_socket_id;
2279 
2280     smc->dr_lmb_enabled = false;
2281     fwc->get_dev_path = spapr_get_fw_dev_path;
2282     nc->nmi_monitor_handler = spapr_nmi;
2283 }
2284 
2285 static const TypeInfo spapr_machine_info = {
2286     .name          = TYPE_SPAPR_MACHINE,
2287     .parent        = TYPE_MACHINE,
2288     .abstract      = true,
2289     .instance_size = sizeof(sPAPRMachineState),
2290     .instance_init = spapr_machine_initfn,
2291     .class_size    = sizeof(sPAPRMachineClass),
2292     .class_init    = spapr_machine_class_init,
2293     .interfaces = (InterfaceInfo[]) {
2294         { TYPE_FW_PATH_PROVIDER },
2295         { TYPE_NMI },
2296         { TYPE_HOTPLUG_HANDLER },
2297         { }
2298     },
2299 };
2300 
2301 #define SPAPR_COMPAT_2_4 \
2302         HW_COMPAT_2_4
2303 
2304 #define SPAPR_COMPAT_2_3 \
2305         SPAPR_COMPAT_2_4 \
2306         HW_COMPAT_2_3 \
2307         {\
2308             .driver   = "spapr-pci-host-bridge",\
2309             .property = "dynamic-reconfiguration",\
2310             .value    = "off",\
2311         },
2312 
2313 #define SPAPR_COMPAT_2_2 \
2314         SPAPR_COMPAT_2_3 \
2315         HW_COMPAT_2_2 \
2316         {\
2317             .driver   = TYPE_SPAPR_PCI_HOST_BRIDGE,\
2318             .property = "mem_win_size",\
2319             .value    = "0x20000000",\
2320         },
2321 
2322 #define SPAPR_COMPAT_2_1 \
2323         SPAPR_COMPAT_2_2 \
2324         HW_COMPAT_2_1
2325 
2326 static void spapr_compat_2_3(Object *obj)
2327 {
2328     savevm_skip_section_footers();
2329     global_state_set_optional();
2330 }
2331 
2332 static void spapr_compat_2_2(Object *obj)
2333 {
2334     spapr_compat_2_3(obj);
2335 }
2336 
2337 static void spapr_compat_2_1(Object *obj)
2338 {
2339     spapr_compat_2_2(obj);
2340 }
2341 
2342 static void spapr_machine_2_3_instance_init(Object *obj)
2343 {
2344     spapr_compat_2_3(obj);
2345     spapr_machine_initfn(obj);
2346 }
2347 
2348 static void spapr_machine_2_2_instance_init(Object *obj)
2349 {
2350     spapr_compat_2_2(obj);
2351     spapr_machine_initfn(obj);
2352 }
2353 
2354 static void spapr_machine_2_1_instance_init(Object *obj)
2355 {
2356     spapr_compat_2_1(obj);
2357     spapr_machine_initfn(obj);
2358 }
2359 
2360 static void spapr_machine_2_1_class_init(ObjectClass *oc, void *data)
2361 {
2362     MachineClass *mc = MACHINE_CLASS(oc);
2363     static GlobalProperty compat_props[] = {
2364         SPAPR_COMPAT_2_1
2365         { /* end of list */ }
2366     };
2367 
2368     mc->desc = "pSeries Logical Partition (PAPR compliant) v2.1";
2369     mc->compat_props = compat_props;
2370 }
2371 
2372 static const TypeInfo spapr_machine_2_1_info = {
2373     .name          = MACHINE_TYPE_NAME("pseries-2.1"),
2374     .parent        = TYPE_SPAPR_MACHINE,
2375     .class_init    = spapr_machine_2_1_class_init,
2376     .instance_init = spapr_machine_2_1_instance_init,
2377 };
2378 
2379 static void spapr_machine_2_2_class_init(ObjectClass *oc, void *data)
2380 {
2381     static GlobalProperty compat_props[] = {
2382         SPAPR_COMPAT_2_2
2383         { /* end of list */ }
2384     };
2385     MachineClass *mc = MACHINE_CLASS(oc);
2386 
2387     mc->desc = "pSeries Logical Partition (PAPR compliant) v2.2";
2388     mc->compat_props = compat_props;
2389 }
2390 
2391 static const TypeInfo spapr_machine_2_2_info = {
2392     .name          = MACHINE_TYPE_NAME("pseries-2.2"),
2393     .parent        = TYPE_SPAPR_MACHINE,
2394     .class_init    = spapr_machine_2_2_class_init,
2395     .instance_init = spapr_machine_2_2_instance_init,
2396 };
2397 
2398 static void spapr_machine_2_3_class_init(ObjectClass *oc, void *data)
2399 {
2400     static GlobalProperty compat_props[] = {
2401         SPAPR_COMPAT_2_3
2402         { /* end of list */ }
2403     };
2404     MachineClass *mc = MACHINE_CLASS(oc);
2405 
2406     mc->desc = "pSeries Logical Partition (PAPR compliant) v2.3";
2407     mc->compat_props = compat_props;
2408 }
2409 
2410 static const TypeInfo spapr_machine_2_3_info = {
2411     .name          = MACHINE_TYPE_NAME("pseries-2.3"),
2412     .parent        = TYPE_SPAPR_MACHINE,
2413     .class_init    = spapr_machine_2_3_class_init,
2414     .instance_init = spapr_machine_2_3_instance_init,
2415 };
2416 
2417 static void spapr_machine_2_4_class_init(ObjectClass *oc, void *data)
2418 {
2419     static GlobalProperty compat_props[] = {
2420         SPAPR_COMPAT_2_4
2421         { /* end of list */ }
2422     };
2423     MachineClass *mc = MACHINE_CLASS(oc);
2424 
2425     mc->desc = "pSeries Logical Partition (PAPR compliant) v2.4";
2426     mc->compat_props = compat_props;
2427 }
2428 
2429 static const TypeInfo spapr_machine_2_4_info = {
2430     .name          = MACHINE_TYPE_NAME("pseries-2.4"),
2431     .parent        = TYPE_SPAPR_MACHINE,
2432     .class_init    = spapr_machine_2_4_class_init,
2433 };
2434 
2435 static void spapr_machine_2_5_class_init(ObjectClass *oc, void *data)
2436 {
2437     MachineClass *mc = MACHINE_CLASS(oc);
2438     sPAPRMachineClass *smc = SPAPR_MACHINE_CLASS(oc);
2439 
2440     mc->name = "pseries-2.5";
2441     mc->desc = "pSeries Logical Partition (PAPR compliant) v2.5";
2442     mc->alias = "pseries";
2443     mc->is_default = 1;
2444     smc->dr_lmb_enabled = true;
2445 }
2446 
2447 static const TypeInfo spapr_machine_2_5_info = {
2448     .name          = MACHINE_TYPE_NAME("pseries-2.5"),
2449     .parent        = TYPE_SPAPR_MACHINE,
2450     .class_init    = spapr_machine_2_5_class_init,
2451 };
2452 
2453 static void spapr_machine_register_types(void)
2454 {
2455     type_register_static(&spapr_machine_info);
2456     type_register_static(&spapr_machine_2_1_info);
2457     type_register_static(&spapr_machine_2_2_info);
2458     type_register_static(&spapr_machine_2_3_info);
2459     type_register_static(&spapr_machine_2_4_info);
2460     type_register_static(&spapr_machine_2_5_info);
2461 }
2462 
2463 type_init(spapr_machine_register_types)
2464