xref: /openbmc/qemu/hw/ppc/spapr.c (revision 9884abee)
1 /*
2  * QEMU PowerPC pSeries Logical Partition (aka sPAPR) hardware System Emulator
3  *
4  * Copyright (c) 2004-2007 Fabrice Bellard
5  * Copyright (c) 2007 Jocelyn Mayer
6  * Copyright (c) 2010 David Gibson, IBM Corporation.
7  *
8  * Permission is hereby granted, free of charge, to any person obtaining a copy
9  * of this software and associated documentation files (the "Software"), to deal
10  * in the Software without restriction, including without limitation the rights
11  * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
12  * copies of the Software, and to permit persons to whom the Software is
13  * furnished to do so, subject to the following conditions:
14  *
15  * The above copyright notice and this permission notice shall be included in
16  * all copies or substantial portions of the Software.
17  *
18  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
19  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
20  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
21  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
22  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
23  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
24  * THE SOFTWARE.
25  *
26  */
27 #include "qemu/osdep.h"
28 #include "sysemu/sysemu.h"
29 #include "sysemu/numa.h"
30 #include "hw/hw.h"
31 #include "hw/fw-path-provider.h"
32 #include "elf.h"
33 #include "net/net.h"
34 #include "sysemu/device_tree.h"
35 #include "sysemu/block-backend.h"
36 #include "sysemu/cpus.h"
37 #include "sysemu/kvm.h"
38 #include "sysemu/device_tree.h"
39 #include "kvm_ppc.h"
40 #include "migration/migration.h"
41 #include "mmu-hash64.h"
42 #include "qom/cpu.h"
43 
44 #include "hw/boards.h"
45 #include "hw/ppc/ppc.h"
46 #include "hw/loader.h"
47 
48 #include "hw/ppc/spapr.h"
49 #include "hw/ppc/spapr_vio.h"
50 #include "hw/pci-host/spapr.h"
51 #include "hw/ppc/xics.h"
52 #include "hw/pci/msi.h"
53 
54 #include "hw/pci/pci.h"
55 #include "hw/scsi/scsi.h"
56 #include "hw/virtio/virtio-scsi.h"
57 
58 #include "exec/address-spaces.h"
59 #include "hw/usb.h"
60 #include "qemu/config-file.h"
61 #include "qemu/error-report.h"
62 #include "trace.h"
63 #include "hw/nmi.h"
64 
65 #include "hw/compat.h"
66 #include "qemu-common.h"
67 
68 #include <libfdt.h>
69 
70 /* SLOF memory layout:
71  *
72  * SLOF raw image loaded at 0, copies its romfs right below the flat
73  * device-tree, then position SLOF itself 31M below that
74  *
75  * So we set FW_OVERHEAD to 40MB which should account for all of that
76  * and more
77  *
78  * We load our kernel at 4M, leaving space for SLOF initial image
79  */
80 #define FDT_MAX_SIZE            0x100000
81 #define RTAS_MAX_SIZE           0x10000
82 #define RTAS_MAX_ADDR           0x80000000 /* RTAS must stay below that */
83 #define FW_MAX_SIZE             0x400000
84 #define FW_FILE_NAME            "slof.bin"
85 #define FW_OVERHEAD             0x2800000
86 #define KERNEL_LOAD_ADDR        FW_MAX_SIZE
87 
88 #define MIN_RMA_SLOF            128UL
89 
90 #define TIMEBASE_FREQ           512000000ULL
91 
92 #define PHANDLE_XICP            0x00001111
93 
94 #define HTAB_SIZE(spapr)        (1ULL << ((spapr)->htab_shift))
95 
96 static XICSState *try_create_xics(const char *type, int nr_servers,
97                                   int nr_irqs, Error **errp)
98 {
99     Error *err = NULL;
100     DeviceState *dev;
101 
102     dev = qdev_create(NULL, type);
103     qdev_prop_set_uint32(dev, "nr_servers", nr_servers);
104     qdev_prop_set_uint32(dev, "nr_irqs", nr_irqs);
105     object_property_set_bool(OBJECT(dev), true, "realized", &err);
106     if (err) {
107         error_propagate(errp, err);
108         object_unparent(OBJECT(dev));
109         return NULL;
110     }
111     return XICS_COMMON(dev);
112 }
113 
114 static XICSState *xics_system_init(MachineState *machine,
115                                    int nr_servers, int nr_irqs)
116 {
117     XICSState *icp = NULL;
118 
119     if (kvm_enabled()) {
120         Error *err = NULL;
121 
122         if (machine_kernel_irqchip_allowed(machine)) {
123             icp = try_create_xics(TYPE_KVM_XICS, nr_servers, nr_irqs, &err);
124         }
125         if (machine_kernel_irqchip_required(machine) && !icp) {
126             error_reportf_err(err,
127                               "kernel_irqchip requested but unavailable: ");
128         } else {
129             error_free(err);
130         }
131     }
132 
133     if (!icp) {
134         icp = try_create_xics(TYPE_XICS, nr_servers, nr_irqs, &error_abort);
135     }
136 
137     return icp;
138 }
139 
140 static int spapr_fixup_cpu_smt_dt(void *fdt, int offset, PowerPCCPU *cpu,
141                                   int smt_threads)
142 {
143     int i, ret = 0;
144     uint32_t servers_prop[smt_threads];
145     uint32_t gservers_prop[smt_threads * 2];
146     int index = ppc_get_vcpu_dt_id(cpu);
147 
148     if (cpu->cpu_version) {
149         ret = fdt_setprop_cell(fdt, offset, "cpu-version", cpu->cpu_version);
150         if (ret < 0) {
151             return ret;
152         }
153     }
154 
155     /* Build interrupt servers and gservers properties */
156     for (i = 0; i < smt_threads; i++) {
157         servers_prop[i] = cpu_to_be32(index + i);
158         /* Hack, direct the group queues back to cpu 0 */
159         gservers_prop[i*2] = cpu_to_be32(index + i);
160         gservers_prop[i*2 + 1] = 0;
161     }
162     ret = fdt_setprop(fdt, offset, "ibm,ppc-interrupt-server#s",
163                       servers_prop, sizeof(servers_prop));
164     if (ret < 0) {
165         return ret;
166     }
167     ret = fdt_setprop(fdt, offset, "ibm,ppc-interrupt-gserver#s",
168                       gservers_prop, sizeof(gservers_prop));
169 
170     return ret;
171 }
172 
173 static int spapr_fixup_cpu_numa_dt(void *fdt, int offset, CPUState *cs)
174 {
175     int ret = 0;
176     PowerPCCPU *cpu = POWERPC_CPU(cs);
177     int index = ppc_get_vcpu_dt_id(cpu);
178     uint32_t associativity[] = {cpu_to_be32(0x5),
179                                 cpu_to_be32(0x0),
180                                 cpu_to_be32(0x0),
181                                 cpu_to_be32(0x0),
182                                 cpu_to_be32(cs->numa_node),
183                                 cpu_to_be32(index)};
184 
185     /* Advertise NUMA via ibm,associativity */
186     if (nb_numa_nodes > 1) {
187         ret = fdt_setprop(fdt, offset, "ibm,associativity", associativity,
188                           sizeof(associativity));
189     }
190 
191     return ret;
192 }
193 
194 static int spapr_fixup_cpu_dt(void *fdt, sPAPRMachineState *spapr)
195 {
196     int ret = 0, offset, cpus_offset;
197     CPUState *cs;
198     char cpu_model[32];
199     int smt = kvmppc_smt_threads();
200     uint32_t pft_size_prop[] = {0, cpu_to_be32(spapr->htab_shift)};
201 
202     CPU_FOREACH(cs) {
203         PowerPCCPU *cpu = POWERPC_CPU(cs);
204         DeviceClass *dc = DEVICE_GET_CLASS(cs);
205         int index = ppc_get_vcpu_dt_id(cpu);
206 
207         if ((index % smt) != 0) {
208             continue;
209         }
210 
211         snprintf(cpu_model, 32, "%s@%x", dc->fw_name, index);
212 
213         cpus_offset = fdt_path_offset(fdt, "/cpus");
214         if (cpus_offset < 0) {
215             cpus_offset = fdt_add_subnode(fdt, fdt_path_offset(fdt, "/"),
216                                           "cpus");
217             if (cpus_offset < 0) {
218                 return cpus_offset;
219             }
220         }
221         offset = fdt_subnode_offset(fdt, cpus_offset, cpu_model);
222         if (offset < 0) {
223             offset = fdt_add_subnode(fdt, cpus_offset, cpu_model);
224             if (offset < 0) {
225                 return offset;
226             }
227         }
228 
229         ret = fdt_setprop(fdt, offset, "ibm,pft-size",
230                           pft_size_prop, sizeof(pft_size_prop));
231         if (ret < 0) {
232             return ret;
233         }
234 
235         ret = spapr_fixup_cpu_numa_dt(fdt, offset, cs);
236         if (ret < 0) {
237             return ret;
238         }
239 
240         ret = spapr_fixup_cpu_smt_dt(fdt, offset, cpu,
241                                      ppc_get_compat_smt_threads(cpu));
242         if (ret < 0) {
243             return ret;
244         }
245     }
246     return ret;
247 }
248 
249 
250 static size_t create_page_sizes_prop(CPUPPCState *env, uint32_t *prop,
251                                      size_t maxsize)
252 {
253     size_t maxcells = maxsize / sizeof(uint32_t);
254     int i, j, count;
255     uint32_t *p = prop;
256 
257     for (i = 0; i < PPC_PAGE_SIZES_MAX_SZ; i++) {
258         struct ppc_one_seg_page_size *sps = &env->sps.sps[i];
259 
260         if (!sps->page_shift) {
261             break;
262         }
263         for (count = 0; count < PPC_PAGE_SIZES_MAX_SZ; count++) {
264             if (sps->enc[count].page_shift == 0) {
265                 break;
266             }
267         }
268         if ((p - prop) >= (maxcells - 3 - count * 2)) {
269             break;
270         }
271         *(p++) = cpu_to_be32(sps->page_shift);
272         *(p++) = cpu_to_be32(sps->slb_enc);
273         *(p++) = cpu_to_be32(count);
274         for (j = 0; j < count; j++) {
275             *(p++) = cpu_to_be32(sps->enc[j].page_shift);
276             *(p++) = cpu_to_be32(sps->enc[j].pte_enc);
277         }
278     }
279 
280     return (p - prop) * sizeof(uint32_t);
281 }
282 
283 static hwaddr spapr_node0_size(void)
284 {
285     MachineState *machine = MACHINE(qdev_get_machine());
286 
287     if (nb_numa_nodes) {
288         int i;
289         for (i = 0; i < nb_numa_nodes; ++i) {
290             if (numa_info[i].node_mem) {
291                 return MIN(pow2floor(numa_info[i].node_mem),
292                            machine->ram_size);
293             }
294         }
295     }
296     return machine->ram_size;
297 }
298 
299 #define _FDT(exp) \
300     do { \
301         int ret = (exp);                                           \
302         if (ret < 0) {                                             \
303             fprintf(stderr, "qemu: error creating device tree: %s: %s\n", \
304                     #exp, fdt_strerror(ret));                      \
305             exit(1);                                               \
306         }                                                          \
307     } while (0)
308 
309 static void add_str(GString *s, const gchar *s1)
310 {
311     g_string_append_len(s, s1, strlen(s1) + 1);
312 }
313 
314 static void *spapr_create_fdt_skel(hwaddr initrd_base,
315                                    hwaddr initrd_size,
316                                    hwaddr kernel_size,
317                                    bool little_endian,
318                                    const char *kernel_cmdline,
319                                    uint32_t epow_irq)
320 {
321     void *fdt;
322     uint32_t start_prop = cpu_to_be32(initrd_base);
323     uint32_t end_prop = cpu_to_be32(initrd_base + initrd_size);
324     GString *hypertas = g_string_sized_new(256);
325     GString *qemu_hypertas = g_string_sized_new(256);
326     uint32_t refpoints[] = {cpu_to_be32(0x4), cpu_to_be32(0x4)};
327     uint32_t interrupt_server_ranges_prop[] = {0, cpu_to_be32(max_cpus)};
328     unsigned char vec5[] = {0x0, 0x0, 0x0, 0x0, 0x0, 0x80};
329     char *buf;
330 
331     add_str(hypertas, "hcall-pft");
332     add_str(hypertas, "hcall-term");
333     add_str(hypertas, "hcall-dabr");
334     add_str(hypertas, "hcall-interrupt");
335     add_str(hypertas, "hcall-tce");
336     add_str(hypertas, "hcall-vio");
337     add_str(hypertas, "hcall-splpar");
338     add_str(hypertas, "hcall-bulk");
339     add_str(hypertas, "hcall-set-mode");
340     add_str(qemu_hypertas, "hcall-memop1");
341 
342     fdt = g_malloc0(FDT_MAX_SIZE);
343     _FDT((fdt_create(fdt, FDT_MAX_SIZE)));
344 
345     if (kernel_size) {
346         _FDT((fdt_add_reservemap_entry(fdt, KERNEL_LOAD_ADDR, kernel_size)));
347     }
348     if (initrd_size) {
349         _FDT((fdt_add_reservemap_entry(fdt, initrd_base, initrd_size)));
350     }
351     _FDT((fdt_finish_reservemap(fdt)));
352 
353     /* Root node */
354     _FDT((fdt_begin_node(fdt, "")));
355     _FDT((fdt_property_string(fdt, "device_type", "chrp")));
356     _FDT((fdt_property_string(fdt, "model", "IBM pSeries (emulated by qemu)")));
357     _FDT((fdt_property_string(fdt, "compatible", "qemu,pseries")));
358 
359     /*
360      * Add info to guest to indentify which host is it being run on
361      * and what is the uuid of the guest
362      */
363     if (kvmppc_get_host_model(&buf)) {
364         _FDT((fdt_property_string(fdt, "host-model", buf)));
365         g_free(buf);
366     }
367     if (kvmppc_get_host_serial(&buf)) {
368         _FDT((fdt_property_string(fdt, "host-serial", buf)));
369         g_free(buf);
370     }
371 
372     buf = g_strdup_printf(UUID_FMT, qemu_uuid[0], qemu_uuid[1],
373                           qemu_uuid[2], qemu_uuid[3], qemu_uuid[4],
374                           qemu_uuid[5], qemu_uuid[6], qemu_uuid[7],
375                           qemu_uuid[8], qemu_uuid[9], qemu_uuid[10],
376                           qemu_uuid[11], qemu_uuid[12], qemu_uuid[13],
377                           qemu_uuid[14], qemu_uuid[15]);
378 
379     _FDT((fdt_property_string(fdt, "vm,uuid", buf)));
380     if (qemu_uuid_set) {
381         _FDT((fdt_property_string(fdt, "system-id", buf)));
382     }
383     g_free(buf);
384 
385     if (qemu_get_vm_name()) {
386         _FDT((fdt_property_string(fdt, "ibm,partition-name",
387                                   qemu_get_vm_name())));
388     }
389 
390     _FDT((fdt_property_cell(fdt, "#address-cells", 0x2)));
391     _FDT((fdt_property_cell(fdt, "#size-cells", 0x2)));
392 
393     /* /chosen */
394     _FDT((fdt_begin_node(fdt, "chosen")));
395 
396     /* Set Form1_affinity */
397     _FDT((fdt_property(fdt, "ibm,architecture-vec-5", vec5, sizeof(vec5))));
398 
399     _FDT((fdt_property_string(fdt, "bootargs", kernel_cmdline)));
400     _FDT((fdt_property(fdt, "linux,initrd-start",
401                        &start_prop, sizeof(start_prop))));
402     _FDT((fdt_property(fdt, "linux,initrd-end",
403                        &end_prop, sizeof(end_prop))));
404     if (kernel_size) {
405         uint64_t kprop[2] = { cpu_to_be64(KERNEL_LOAD_ADDR),
406                               cpu_to_be64(kernel_size) };
407 
408         _FDT((fdt_property(fdt, "qemu,boot-kernel", &kprop, sizeof(kprop))));
409         if (little_endian) {
410             _FDT((fdt_property(fdt, "qemu,boot-kernel-le", NULL, 0)));
411         }
412     }
413     if (boot_menu) {
414         _FDT((fdt_property_cell(fdt, "qemu,boot-menu", boot_menu)));
415     }
416     _FDT((fdt_property_cell(fdt, "qemu,graphic-width", graphic_width)));
417     _FDT((fdt_property_cell(fdt, "qemu,graphic-height", graphic_height)));
418     _FDT((fdt_property_cell(fdt, "qemu,graphic-depth", graphic_depth)));
419 
420     _FDT((fdt_end_node(fdt)));
421 
422     /* RTAS */
423     _FDT((fdt_begin_node(fdt, "rtas")));
424 
425     if (!kvm_enabled() || kvmppc_spapr_use_multitce()) {
426         add_str(hypertas, "hcall-multi-tce");
427     }
428     _FDT((fdt_property(fdt, "ibm,hypertas-functions", hypertas->str,
429                        hypertas->len)));
430     g_string_free(hypertas, TRUE);
431     _FDT((fdt_property(fdt, "qemu,hypertas-functions", qemu_hypertas->str,
432                        qemu_hypertas->len)));
433     g_string_free(qemu_hypertas, TRUE);
434 
435     _FDT((fdt_property(fdt, "ibm,associativity-reference-points",
436         refpoints, sizeof(refpoints))));
437 
438     _FDT((fdt_property_cell(fdt, "rtas-error-log-max", RTAS_ERROR_LOG_MAX)));
439     _FDT((fdt_property_cell(fdt, "rtas-event-scan-rate",
440                             RTAS_EVENT_SCAN_RATE)));
441 
442     if (msi_supported) {
443         _FDT((fdt_property(fdt, "ibm,change-msix-capable", NULL, 0)));
444     }
445 
446     /*
447      * According to PAPR, rtas ibm,os-term does not guarantee a return
448      * back to the guest cpu.
449      *
450      * While an additional ibm,extended-os-term property indicates that
451      * rtas call return will always occur. Set this property.
452      */
453     _FDT((fdt_property(fdt, "ibm,extended-os-term", NULL, 0)));
454 
455     _FDT((fdt_end_node(fdt)));
456 
457     /* interrupt controller */
458     _FDT((fdt_begin_node(fdt, "interrupt-controller")));
459 
460     _FDT((fdt_property_string(fdt, "device_type",
461                               "PowerPC-External-Interrupt-Presentation")));
462     _FDT((fdt_property_string(fdt, "compatible", "IBM,ppc-xicp")));
463     _FDT((fdt_property(fdt, "interrupt-controller", NULL, 0)));
464     _FDT((fdt_property(fdt, "ibm,interrupt-server-ranges",
465                        interrupt_server_ranges_prop,
466                        sizeof(interrupt_server_ranges_prop))));
467     _FDT((fdt_property_cell(fdt, "#interrupt-cells", 2)));
468     _FDT((fdt_property_cell(fdt, "linux,phandle", PHANDLE_XICP)));
469     _FDT((fdt_property_cell(fdt, "phandle", PHANDLE_XICP)));
470 
471     _FDT((fdt_end_node(fdt)));
472 
473     /* vdevice */
474     _FDT((fdt_begin_node(fdt, "vdevice")));
475 
476     _FDT((fdt_property_string(fdt, "device_type", "vdevice")));
477     _FDT((fdt_property_string(fdt, "compatible", "IBM,vdevice")));
478     _FDT((fdt_property_cell(fdt, "#address-cells", 0x1)));
479     _FDT((fdt_property_cell(fdt, "#size-cells", 0x0)));
480     _FDT((fdt_property_cell(fdt, "#interrupt-cells", 0x2)));
481     _FDT((fdt_property(fdt, "interrupt-controller", NULL, 0)));
482 
483     _FDT((fdt_end_node(fdt)));
484 
485     /* event-sources */
486     spapr_events_fdt_skel(fdt, epow_irq);
487 
488     /* /hypervisor node */
489     if (kvm_enabled()) {
490         uint8_t hypercall[16];
491 
492         /* indicate KVM hypercall interface */
493         _FDT((fdt_begin_node(fdt, "hypervisor")));
494         _FDT((fdt_property_string(fdt, "compatible", "linux,kvm")));
495         if (kvmppc_has_cap_fixup_hcalls()) {
496             /*
497              * Older KVM versions with older guest kernels were broken with the
498              * magic page, don't allow the guest to map it.
499              */
500             kvmppc_get_hypercall(first_cpu->env_ptr, hypercall,
501                                  sizeof(hypercall));
502             _FDT((fdt_property(fdt, "hcall-instructions", hypercall,
503                               sizeof(hypercall))));
504         }
505         _FDT((fdt_end_node(fdt)));
506     }
507 
508     _FDT((fdt_end_node(fdt))); /* close root node */
509     _FDT((fdt_finish(fdt)));
510 
511     return fdt;
512 }
513 
514 static int spapr_populate_memory_node(void *fdt, int nodeid, hwaddr start,
515                                        hwaddr size)
516 {
517     uint32_t associativity[] = {
518         cpu_to_be32(0x4), /* length */
519         cpu_to_be32(0x0), cpu_to_be32(0x0),
520         cpu_to_be32(0x0), cpu_to_be32(nodeid)
521     };
522     char mem_name[32];
523     uint64_t mem_reg_property[2];
524     int off;
525 
526     mem_reg_property[0] = cpu_to_be64(start);
527     mem_reg_property[1] = cpu_to_be64(size);
528 
529     sprintf(mem_name, "memory@" TARGET_FMT_lx, start);
530     off = fdt_add_subnode(fdt, 0, mem_name);
531     _FDT(off);
532     _FDT((fdt_setprop_string(fdt, off, "device_type", "memory")));
533     _FDT((fdt_setprop(fdt, off, "reg", mem_reg_property,
534                       sizeof(mem_reg_property))));
535     _FDT((fdt_setprop(fdt, off, "ibm,associativity", associativity,
536                       sizeof(associativity))));
537     return off;
538 }
539 
540 static int spapr_populate_memory(sPAPRMachineState *spapr, void *fdt)
541 {
542     MachineState *machine = MACHINE(spapr);
543     hwaddr mem_start, node_size;
544     int i, nb_nodes = nb_numa_nodes;
545     NodeInfo *nodes = numa_info;
546     NodeInfo ramnode;
547 
548     /* No NUMA nodes, assume there is just one node with whole RAM */
549     if (!nb_numa_nodes) {
550         nb_nodes = 1;
551         ramnode.node_mem = machine->ram_size;
552         nodes = &ramnode;
553     }
554 
555     for (i = 0, mem_start = 0; i < nb_nodes; ++i) {
556         if (!nodes[i].node_mem) {
557             continue;
558         }
559         if (mem_start >= machine->ram_size) {
560             node_size = 0;
561         } else {
562             node_size = nodes[i].node_mem;
563             if (node_size > machine->ram_size - mem_start) {
564                 node_size = machine->ram_size - mem_start;
565             }
566         }
567         if (!mem_start) {
568             /* ppc_spapr_init() checks for rma_size <= node0_size already */
569             spapr_populate_memory_node(fdt, i, 0, spapr->rma_size);
570             mem_start += spapr->rma_size;
571             node_size -= spapr->rma_size;
572         }
573         for ( ; node_size; ) {
574             hwaddr sizetmp = pow2floor(node_size);
575 
576             /* mem_start != 0 here */
577             if (ctzl(mem_start) < ctzl(sizetmp)) {
578                 sizetmp = 1ULL << ctzl(mem_start);
579             }
580 
581             spapr_populate_memory_node(fdt, i, mem_start, sizetmp);
582             node_size -= sizetmp;
583             mem_start += sizetmp;
584         }
585     }
586 
587     return 0;
588 }
589 
590 static void spapr_populate_cpu_dt(CPUState *cs, void *fdt, int offset,
591                                   sPAPRMachineState *spapr)
592 {
593     PowerPCCPU *cpu = POWERPC_CPU(cs);
594     CPUPPCState *env = &cpu->env;
595     PowerPCCPUClass *pcc = POWERPC_CPU_GET_CLASS(cs);
596     int index = ppc_get_vcpu_dt_id(cpu);
597     uint32_t segs[] = {cpu_to_be32(28), cpu_to_be32(40),
598                        0xffffffff, 0xffffffff};
599     uint32_t tbfreq = kvm_enabled() ? kvmppc_get_tbfreq() : TIMEBASE_FREQ;
600     uint32_t cpufreq = kvm_enabled() ? kvmppc_get_clockfreq() : 1000000000;
601     uint32_t page_sizes_prop[64];
602     size_t page_sizes_prop_size;
603     uint32_t vcpus_per_socket = smp_threads * smp_cores;
604     uint32_t pft_size_prop[] = {0, cpu_to_be32(spapr->htab_shift)};
605 
606     /* Note: we keep CI large pages off for now because a 64K capable guest
607      * provisioned with large pages might otherwise try to map a qemu
608      * framebuffer (or other kind of memory mapped PCI BAR) using 64K pages
609      * even if that qemu runs on a 4k host.
610      *
611      * We can later add this bit back when we are confident this is not
612      * an issue (!HV KVM or 64K host)
613      */
614     uint8_t pa_features_206[] = { 6, 0,
615         0xf6, 0x1f, 0xc7, 0x00, 0x80, 0xc0 };
616     uint8_t pa_features_207[] = { 24, 0,
617         0xf6, 0x1f, 0xc7, 0xc0, 0x80, 0xf0,
618         0x80, 0x00, 0x00, 0x00, 0x00, 0x00,
619         0x00, 0x00, 0x00, 0x00, 0x80, 0x00,
620         0x80, 0x00, 0x80, 0x00, 0x80, 0x00 };
621     uint8_t *pa_features;
622     size_t pa_size;
623 
624     _FDT((fdt_setprop_cell(fdt, offset, "reg", index)));
625     _FDT((fdt_setprop_string(fdt, offset, "device_type", "cpu")));
626 
627     _FDT((fdt_setprop_cell(fdt, offset, "cpu-version", env->spr[SPR_PVR])));
628     _FDT((fdt_setprop_cell(fdt, offset, "d-cache-block-size",
629                            env->dcache_line_size)));
630     _FDT((fdt_setprop_cell(fdt, offset, "d-cache-line-size",
631                            env->dcache_line_size)));
632     _FDT((fdt_setprop_cell(fdt, offset, "i-cache-block-size",
633                            env->icache_line_size)));
634     _FDT((fdt_setprop_cell(fdt, offset, "i-cache-line-size",
635                            env->icache_line_size)));
636 
637     if (pcc->l1_dcache_size) {
638         _FDT((fdt_setprop_cell(fdt, offset, "d-cache-size",
639                                pcc->l1_dcache_size)));
640     } else {
641         fprintf(stderr, "Warning: Unknown L1 dcache size for cpu\n");
642     }
643     if (pcc->l1_icache_size) {
644         _FDT((fdt_setprop_cell(fdt, offset, "i-cache-size",
645                                pcc->l1_icache_size)));
646     } else {
647         fprintf(stderr, "Warning: Unknown L1 icache size for cpu\n");
648     }
649 
650     _FDT((fdt_setprop_cell(fdt, offset, "timebase-frequency", tbfreq)));
651     _FDT((fdt_setprop_cell(fdt, offset, "clock-frequency", cpufreq)));
652     _FDT((fdt_setprop_cell(fdt, offset, "slb-size", env->slb_nr)));
653     _FDT((fdt_setprop_cell(fdt, offset, "ibm,slb-size", env->slb_nr)));
654     _FDT((fdt_setprop_string(fdt, offset, "status", "okay")));
655     _FDT((fdt_setprop(fdt, offset, "64-bit", NULL, 0)));
656 
657     if (env->spr_cb[SPR_PURR].oea_read) {
658         _FDT((fdt_setprop(fdt, offset, "ibm,purr", NULL, 0)));
659     }
660 
661     if (env->mmu_model & POWERPC_MMU_1TSEG) {
662         _FDT((fdt_setprop(fdt, offset, "ibm,processor-segment-sizes",
663                           segs, sizeof(segs))));
664     }
665 
666     /* Advertise VMX/VSX (vector extensions) if available
667      *   0 / no property == no vector extensions
668      *   1               == VMX / Altivec available
669      *   2               == VSX available */
670     if (env->insns_flags & PPC_ALTIVEC) {
671         uint32_t vmx = (env->insns_flags2 & PPC2_VSX) ? 2 : 1;
672 
673         _FDT((fdt_setprop_cell(fdt, offset, "ibm,vmx", vmx)));
674     }
675 
676     /* Advertise DFP (Decimal Floating Point) if available
677      *   0 / no property == no DFP
678      *   1               == DFP available */
679     if (env->insns_flags2 & PPC2_DFP) {
680         _FDT((fdt_setprop_cell(fdt, offset, "ibm,dfp", 1)));
681     }
682 
683     page_sizes_prop_size = create_page_sizes_prop(env, page_sizes_prop,
684                                                   sizeof(page_sizes_prop));
685     if (page_sizes_prop_size) {
686         _FDT((fdt_setprop(fdt, offset, "ibm,segment-page-sizes",
687                           page_sizes_prop, page_sizes_prop_size)));
688     }
689 
690     /* Do the ibm,pa-features property, adjust it for ci-large-pages */
691     if (env->mmu_model == POWERPC_MMU_2_06) {
692         pa_features = pa_features_206;
693         pa_size = sizeof(pa_features_206);
694     } else /* env->mmu_model == POWERPC_MMU_2_07 */ {
695         pa_features = pa_features_207;
696         pa_size = sizeof(pa_features_207);
697     }
698     if (env->ci_large_pages) {
699         pa_features[3] |= 0x20;
700     }
701     _FDT((fdt_setprop(fdt, offset, "ibm,pa-features", pa_features, pa_size)));
702 
703     _FDT((fdt_setprop_cell(fdt, offset, "ibm,chip-id",
704                            cs->cpu_index / vcpus_per_socket)));
705 
706     _FDT((fdt_setprop(fdt, offset, "ibm,pft-size",
707                       pft_size_prop, sizeof(pft_size_prop))));
708 
709     _FDT(spapr_fixup_cpu_numa_dt(fdt, offset, cs));
710 
711     _FDT(spapr_fixup_cpu_smt_dt(fdt, offset, cpu,
712                                 ppc_get_compat_smt_threads(cpu)));
713 }
714 
715 static void spapr_populate_cpus_dt_node(void *fdt, sPAPRMachineState *spapr)
716 {
717     CPUState *cs;
718     int cpus_offset;
719     char *nodename;
720     int smt = kvmppc_smt_threads();
721 
722     cpus_offset = fdt_add_subnode(fdt, 0, "cpus");
723     _FDT(cpus_offset);
724     _FDT((fdt_setprop_cell(fdt, cpus_offset, "#address-cells", 0x1)));
725     _FDT((fdt_setprop_cell(fdt, cpus_offset, "#size-cells", 0x0)));
726 
727     /*
728      * We walk the CPUs in reverse order to ensure that CPU DT nodes
729      * created by fdt_add_subnode() end up in the right order in FDT
730      * for the guest kernel the enumerate the CPUs correctly.
731      */
732     CPU_FOREACH_REVERSE(cs) {
733         PowerPCCPU *cpu = POWERPC_CPU(cs);
734         int index = ppc_get_vcpu_dt_id(cpu);
735         DeviceClass *dc = DEVICE_GET_CLASS(cs);
736         int offset;
737 
738         if ((index % smt) != 0) {
739             continue;
740         }
741 
742         nodename = g_strdup_printf("%s@%x", dc->fw_name, index);
743         offset = fdt_add_subnode(fdt, cpus_offset, nodename);
744         g_free(nodename);
745         _FDT(offset);
746         spapr_populate_cpu_dt(cs, fdt, offset, spapr);
747     }
748 
749 }
750 
751 /*
752  * Adds ibm,dynamic-reconfiguration-memory node.
753  * Refer to docs/specs/ppc-spapr-hotplug.txt for the documentation
754  * of this device tree node.
755  */
756 static int spapr_populate_drconf_memory(sPAPRMachineState *spapr, void *fdt)
757 {
758     MachineState *machine = MACHINE(spapr);
759     int ret, i, offset;
760     uint64_t lmb_size = SPAPR_MEMORY_BLOCK_SIZE;
761     uint32_t prop_lmb_size[] = {0, cpu_to_be32(lmb_size)};
762     uint32_t nr_lmbs = (machine->maxram_size - machine->ram_size)/lmb_size;
763     uint32_t *int_buf, *cur_index, buf_len;
764     int nr_nodes = nb_numa_nodes ? nb_numa_nodes : 1;
765 
766     /*
767      * Allocate enough buffer size to fit in ibm,dynamic-memory
768      * or ibm,associativity-lookup-arrays
769      */
770     buf_len = MAX(nr_lmbs * SPAPR_DR_LMB_LIST_ENTRY_SIZE + 1, nr_nodes * 4 + 2)
771               * sizeof(uint32_t);
772     cur_index = int_buf = g_malloc0(buf_len);
773 
774     offset = fdt_add_subnode(fdt, 0, "ibm,dynamic-reconfiguration-memory");
775 
776     ret = fdt_setprop(fdt, offset, "ibm,lmb-size", prop_lmb_size,
777                     sizeof(prop_lmb_size));
778     if (ret < 0) {
779         goto out;
780     }
781 
782     ret = fdt_setprop_cell(fdt, offset, "ibm,memory-flags-mask", 0xff);
783     if (ret < 0) {
784         goto out;
785     }
786 
787     ret = fdt_setprop_cell(fdt, offset, "ibm,memory-preservation-time", 0x0);
788     if (ret < 0) {
789         goto out;
790     }
791 
792     /* ibm,dynamic-memory */
793     int_buf[0] = cpu_to_be32(nr_lmbs);
794     cur_index++;
795     for (i = 0; i < nr_lmbs; i++) {
796         sPAPRDRConnector *drc;
797         sPAPRDRConnectorClass *drck;
798         uint64_t addr = i * lmb_size + spapr->hotplug_memory.base;;
799         uint32_t *dynamic_memory = cur_index;
800 
801         drc = spapr_dr_connector_by_id(SPAPR_DR_CONNECTOR_TYPE_LMB,
802                                        addr/lmb_size);
803         g_assert(drc);
804         drck = SPAPR_DR_CONNECTOR_GET_CLASS(drc);
805 
806         dynamic_memory[0] = cpu_to_be32(addr >> 32);
807         dynamic_memory[1] = cpu_to_be32(addr & 0xffffffff);
808         dynamic_memory[2] = cpu_to_be32(drck->get_index(drc));
809         dynamic_memory[3] = cpu_to_be32(0); /* reserved */
810         dynamic_memory[4] = cpu_to_be32(numa_get_node(addr, NULL));
811         if (addr < machine->ram_size ||
812                     memory_region_present(get_system_memory(), addr)) {
813             dynamic_memory[5] = cpu_to_be32(SPAPR_LMB_FLAGS_ASSIGNED);
814         } else {
815             dynamic_memory[5] = cpu_to_be32(0);
816         }
817 
818         cur_index += SPAPR_DR_LMB_LIST_ENTRY_SIZE;
819     }
820     ret = fdt_setprop(fdt, offset, "ibm,dynamic-memory", int_buf, buf_len);
821     if (ret < 0) {
822         goto out;
823     }
824 
825     /* ibm,associativity-lookup-arrays */
826     cur_index = int_buf;
827     int_buf[0] = cpu_to_be32(nr_nodes);
828     int_buf[1] = cpu_to_be32(4); /* Number of entries per associativity list */
829     cur_index += 2;
830     for (i = 0; i < nr_nodes; i++) {
831         uint32_t associativity[] = {
832             cpu_to_be32(0x0),
833             cpu_to_be32(0x0),
834             cpu_to_be32(0x0),
835             cpu_to_be32(i)
836         };
837         memcpy(cur_index, associativity, sizeof(associativity));
838         cur_index += 4;
839     }
840     ret = fdt_setprop(fdt, offset, "ibm,associativity-lookup-arrays", int_buf,
841             (cur_index - int_buf) * sizeof(uint32_t));
842 out:
843     g_free(int_buf);
844     return ret;
845 }
846 
847 int spapr_h_cas_compose_response(sPAPRMachineState *spapr,
848                                  target_ulong addr, target_ulong size,
849                                  bool cpu_update, bool memory_update)
850 {
851     void *fdt, *fdt_skel;
852     sPAPRDeviceTreeUpdateHeader hdr = { .version_id = 1 };
853     sPAPRMachineClass *smc = SPAPR_MACHINE_GET_CLASS(qdev_get_machine());
854 
855     size -= sizeof(hdr);
856 
857     /* Create sceleton */
858     fdt_skel = g_malloc0(size);
859     _FDT((fdt_create(fdt_skel, size)));
860     _FDT((fdt_begin_node(fdt_skel, "")));
861     _FDT((fdt_end_node(fdt_skel)));
862     _FDT((fdt_finish(fdt_skel)));
863     fdt = g_malloc0(size);
864     _FDT((fdt_open_into(fdt_skel, fdt, size)));
865     g_free(fdt_skel);
866 
867     /* Fixup cpu nodes */
868     if (cpu_update) {
869         _FDT((spapr_fixup_cpu_dt(fdt, spapr)));
870     }
871 
872     /* Generate memory nodes or ibm,dynamic-reconfiguration-memory node */
873     if (memory_update && smc->dr_lmb_enabled) {
874         _FDT((spapr_populate_drconf_memory(spapr, fdt)));
875     }
876 
877     /* Pack resulting tree */
878     _FDT((fdt_pack(fdt)));
879 
880     if (fdt_totalsize(fdt) + sizeof(hdr) > size) {
881         trace_spapr_cas_failed(size);
882         return -1;
883     }
884 
885     cpu_physical_memory_write(addr, &hdr, sizeof(hdr));
886     cpu_physical_memory_write(addr + sizeof(hdr), fdt, fdt_totalsize(fdt));
887     trace_spapr_cas_continue(fdt_totalsize(fdt) + sizeof(hdr));
888     g_free(fdt);
889 
890     return 0;
891 }
892 
893 static void spapr_finalize_fdt(sPAPRMachineState *spapr,
894                                hwaddr fdt_addr,
895                                hwaddr rtas_addr,
896                                hwaddr rtas_size)
897 {
898     MachineState *machine = MACHINE(qdev_get_machine());
899     sPAPRMachineClass *smc = SPAPR_MACHINE_GET_CLASS(machine);
900     const char *boot_device = machine->boot_order;
901     int ret, i;
902     size_t cb = 0;
903     char *bootlist;
904     void *fdt;
905     sPAPRPHBState *phb;
906 
907     fdt = g_malloc(FDT_MAX_SIZE);
908 
909     /* open out the base tree into a temp buffer for the final tweaks */
910     _FDT((fdt_open_into(spapr->fdt_skel, fdt, FDT_MAX_SIZE)));
911 
912     ret = spapr_populate_memory(spapr, fdt);
913     if (ret < 0) {
914         fprintf(stderr, "couldn't setup memory nodes in fdt\n");
915         exit(1);
916     }
917 
918     ret = spapr_populate_vdevice(spapr->vio_bus, fdt);
919     if (ret < 0) {
920         fprintf(stderr, "couldn't setup vio devices in fdt\n");
921         exit(1);
922     }
923 
924     if (object_resolve_path_type("", TYPE_SPAPR_RNG, NULL)) {
925         ret = spapr_rng_populate_dt(fdt);
926         if (ret < 0) {
927             fprintf(stderr, "could not set up rng device in the fdt\n");
928             exit(1);
929         }
930     }
931 
932     QLIST_FOREACH(phb, &spapr->phbs, list) {
933         ret = spapr_populate_pci_dt(phb, PHANDLE_XICP, fdt);
934     }
935 
936     if (ret < 0) {
937         fprintf(stderr, "couldn't setup PCI devices in fdt\n");
938         exit(1);
939     }
940 
941     /* RTAS */
942     ret = spapr_rtas_device_tree_setup(fdt, rtas_addr, rtas_size);
943     if (ret < 0) {
944         fprintf(stderr, "Couldn't set up RTAS device tree properties\n");
945     }
946 
947     /* cpus */
948     spapr_populate_cpus_dt_node(fdt, spapr);
949 
950     bootlist = get_boot_devices_list(&cb, true);
951     if (cb && bootlist) {
952         int offset = fdt_path_offset(fdt, "/chosen");
953         if (offset < 0) {
954             exit(1);
955         }
956         for (i = 0; i < cb; i++) {
957             if (bootlist[i] == '\n') {
958                 bootlist[i] = ' ';
959             }
960 
961         }
962         ret = fdt_setprop_string(fdt, offset, "qemu,boot-list", bootlist);
963     }
964 
965     if (boot_device && strlen(boot_device)) {
966         int offset = fdt_path_offset(fdt, "/chosen");
967 
968         if (offset < 0) {
969             exit(1);
970         }
971         fdt_setprop_string(fdt, offset, "qemu,boot-device", boot_device);
972     }
973 
974     if (!spapr->has_graphics) {
975         spapr_populate_chosen_stdout(fdt, spapr->vio_bus);
976     }
977 
978     if (smc->dr_lmb_enabled) {
979         _FDT(spapr_drc_populate_dt(fdt, 0, NULL, SPAPR_DR_CONNECTOR_TYPE_LMB));
980     }
981 
982     _FDT((fdt_pack(fdt)));
983 
984     if (fdt_totalsize(fdt) > FDT_MAX_SIZE) {
985         error_report("FDT too big ! 0x%x bytes (max is 0x%x)",
986                      fdt_totalsize(fdt), FDT_MAX_SIZE);
987         exit(1);
988     }
989 
990     qemu_fdt_dumpdtb(fdt, fdt_totalsize(fdt));
991     cpu_physical_memory_write(fdt_addr, fdt, fdt_totalsize(fdt));
992 
993     g_free(bootlist);
994     g_free(fdt);
995 }
996 
997 static uint64_t translate_kernel_address(void *opaque, uint64_t addr)
998 {
999     return (addr & 0x0fffffff) + KERNEL_LOAD_ADDR;
1000 }
1001 
1002 static void emulate_spapr_hypercall(PowerPCCPU *cpu)
1003 {
1004     CPUPPCState *env = &cpu->env;
1005 
1006     if (msr_pr) {
1007         hcall_dprintf("Hypercall made with MSR[PR]=1\n");
1008         env->gpr[3] = H_PRIVILEGE;
1009     } else {
1010         env->gpr[3] = spapr_hypercall(cpu, env->gpr[3], &env->gpr[4]);
1011     }
1012 }
1013 
1014 #define HPTE(_table, _i)   (void *)(((uint64_t *)(_table)) + ((_i) * 2))
1015 #define HPTE_VALID(_hpte)  (tswap64(*((uint64_t *)(_hpte))) & HPTE64_V_VALID)
1016 #define HPTE_DIRTY(_hpte)  (tswap64(*((uint64_t *)(_hpte))) & HPTE64_V_HPTE_DIRTY)
1017 #define CLEAN_HPTE(_hpte)  ((*(uint64_t *)(_hpte)) &= tswap64(~HPTE64_V_HPTE_DIRTY))
1018 #define DIRTY_HPTE(_hpte)  ((*(uint64_t *)(_hpte)) |= tswap64(HPTE64_V_HPTE_DIRTY))
1019 
1020 static void spapr_alloc_htab(sPAPRMachineState *spapr)
1021 {
1022     long shift;
1023     int index;
1024 
1025     /* allocate hash page table.  For now we always make this 16mb,
1026      * later we should probably make it scale to the size of guest
1027      * RAM */
1028 
1029     shift = kvmppc_reset_htab(spapr->htab_shift);
1030     if (shift < 0) {
1031         /*
1032          * For HV KVM, host kernel will return -ENOMEM when requested
1033          * HTAB size can't be allocated.
1034          */
1035         error_setg(&error_abort, "Failed to allocate HTAB of requested size, try with smaller maxmem");
1036     } else if (shift > 0) {
1037         /*
1038          * Kernel handles htab, we don't need to allocate one
1039          *
1040          * Older kernels can fall back to lower HTAB shift values,
1041          * but we don't allow booting of such guests.
1042          */
1043         if (shift != spapr->htab_shift) {
1044             error_setg(&error_abort, "Failed to allocate HTAB of requested size, try with smaller maxmem");
1045         }
1046 
1047         spapr->htab_shift = shift;
1048         kvmppc_kern_htab = true;
1049     } else {
1050         /* Allocate htab */
1051         spapr->htab = qemu_memalign(HTAB_SIZE(spapr), HTAB_SIZE(spapr));
1052 
1053         /* And clear it */
1054         memset(spapr->htab, 0, HTAB_SIZE(spapr));
1055 
1056         for (index = 0; index < HTAB_SIZE(spapr) / HASH_PTE_SIZE_64; index++) {
1057             DIRTY_HPTE(HPTE(spapr->htab, index));
1058         }
1059     }
1060 }
1061 
1062 /*
1063  * Clear HTAB entries during reset.
1064  *
1065  * If host kernel has allocated HTAB, KVM_PPC_ALLOCATE_HTAB ioctl is
1066  * used to clear HTAB. Otherwise QEMU-allocated HTAB is cleared manually.
1067  */
1068 static void spapr_reset_htab(sPAPRMachineState *spapr)
1069 {
1070     long shift;
1071     int index;
1072 
1073     shift = kvmppc_reset_htab(spapr->htab_shift);
1074     if (shift < 0) {
1075         error_setg(&error_abort, "Failed to reset HTAB");
1076     } else if (shift > 0) {
1077         if (shift != spapr->htab_shift) {
1078             error_setg(&error_abort, "Requested HTAB allocation failed during reset");
1079         }
1080 
1081         /* Tell readers to update their file descriptor */
1082         if (spapr->htab_fd >= 0) {
1083             spapr->htab_fd_stale = true;
1084         }
1085     } else {
1086         memset(spapr->htab, 0, HTAB_SIZE(spapr));
1087 
1088         for (index = 0; index < HTAB_SIZE(spapr) / HASH_PTE_SIZE_64; index++) {
1089             DIRTY_HPTE(HPTE(spapr->htab, index));
1090         }
1091     }
1092 
1093     /* Update the RMA size if necessary */
1094     if (spapr->vrma_adjust) {
1095         spapr->rma_size = kvmppc_rma_size(spapr_node0_size(),
1096                                           spapr->htab_shift);
1097     }
1098 }
1099 
1100 static int find_unknown_sysbus_device(SysBusDevice *sbdev, void *opaque)
1101 {
1102     bool matched = false;
1103 
1104     if (object_dynamic_cast(OBJECT(sbdev), TYPE_SPAPR_PCI_HOST_BRIDGE)) {
1105         matched = true;
1106     }
1107 
1108     if (!matched) {
1109         error_report("Device %s is not supported by this machine yet.",
1110                      qdev_fw_name(DEVICE(sbdev)));
1111         exit(1);
1112     }
1113 
1114     return 0;
1115 }
1116 
1117 /*
1118  * A guest reset will cause spapr->htab_fd to become stale if being used.
1119  * Reopen the file descriptor to make sure the whole HTAB is properly read.
1120  */
1121 static int spapr_check_htab_fd(sPAPRMachineState *spapr)
1122 {
1123     int rc = 0;
1124 
1125     if (spapr->htab_fd_stale) {
1126         close(spapr->htab_fd);
1127         spapr->htab_fd = kvmppc_get_htab_fd(false);
1128         if (spapr->htab_fd < 0) {
1129             error_report("Unable to open fd for reading hash table from KVM: "
1130                          "%s", strerror(errno));
1131             rc = -1;
1132         }
1133         spapr->htab_fd_stale = false;
1134     }
1135 
1136     return rc;
1137 }
1138 
1139 static void ppc_spapr_reset(void)
1140 {
1141     sPAPRMachineState *spapr = SPAPR_MACHINE(qdev_get_machine());
1142     PowerPCCPU *first_ppc_cpu;
1143     uint32_t rtas_limit;
1144 
1145     /* Check for unknown sysbus devices */
1146     foreach_dynamic_sysbus_device(find_unknown_sysbus_device, NULL);
1147 
1148     /* Reset the hash table & recalc the RMA */
1149     spapr_reset_htab(spapr);
1150 
1151     qemu_devices_reset();
1152 
1153     /*
1154      * We place the device tree and RTAS just below either the top of the RMA,
1155      * or just below 2GB, whichever is lowere, so that it can be
1156      * processed with 32-bit real mode code if necessary
1157      */
1158     rtas_limit = MIN(spapr->rma_size, RTAS_MAX_ADDR);
1159     spapr->rtas_addr = rtas_limit - RTAS_MAX_SIZE;
1160     spapr->fdt_addr = spapr->rtas_addr - FDT_MAX_SIZE;
1161 
1162     /* Load the fdt */
1163     spapr_finalize_fdt(spapr, spapr->fdt_addr, spapr->rtas_addr,
1164                        spapr->rtas_size);
1165 
1166     /* Copy RTAS over */
1167     cpu_physical_memory_write(spapr->rtas_addr, spapr->rtas_blob,
1168                               spapr->rtas_size);
1169 
1170     /* Set up the entry state */
1171     first_ppc_cpu = POWERPC_CPU(first_cpu);
1172     first_ppc_cpu->env.gpr[3] = spapr->fdt_addr;
1173     first_ppc_cpu->env.gpr[5] = 0;
1174     first_cpu->halted = 0;
1175     first_ppc_cpu->env.nip = SPAPR_ENTRY_POINT;
1176 
1177 }
1178 
1179 static void spapr_cpu_reset(void *opaque)
1180 {
1181     sPAPRMachineState *spapr = SPAPR_MACHINE(qdev_get_machine());
1182     PowerPCCPU *cpu = opaque;
1183     CPUState *cs = CPU(cpu);
1184     CPUPPCState *env = &cpu->env;
1185 
1186     cpu_reset(cs);
1187 
1188     /* All CPUs start halted.  CPU0 is unhalted from the machine level
1189      * reset code and the rest are explicitly started up by the guest
1190      * using an RTAS call */
1191     cs->halted = 1;
1192 
1193     env->spr[SPR_HIOR] = 0;
1194 
1195     env->external_htab = (uint8_t *)spapr->htab;
1196     if (kvm_enabled() && !env->external_htab) {
1197         /*
1198          * HV KVM, set external_htab to 1 so our ppc_hash64_load_hpte*
1199          * functions do the right thing.
1200          */
1201         env->external_htab = (void *)1;
1202     }
1203     env->htab_base = -1;
1204     /*
1205      * htab_mask is the mask used to normalize hash value to PTEG index.
1206      * htab_shift is log2 of hash table size.
1207      * We have 8 hpte per group, and each hpte is 16 bytes.
1208      * ie have 128 bytes per hpte entry.
1209      */
1210     env->htab_mask = (1ULL << (spapr->htab_shift - 7)) - 1;
1211     env->spr[SPR_SDR1] = (target_ulong)(uintptr_t)spapr->htab |
1212         (spapr->htab_shift - 18);
1213 }
1214 
1215 static void spapr_create_nvram(sPAPRMachineState *spapr)
1216 {
1217     DeviceState *dev = qdev_create(&spapr->vio_bus->bus, "spapr-nvram");
1218     DriveInfo *dinfo = drive_get(IF_PFLASH, 0, 0);
1219 
1220     if (dinfo) {
1221         qdev_prop_set_drive(dev, "drive", blk_by_legacy_dinfo(dinfo),
1222                             &error_fatal);
1223     }
1224 
1225     qdev_init_nofail(dev);
1226 
1227     spapr->nvram = (struct sPAPRNVRAM *)dev;
1228 }
1229 
1230 static void spapr_rtc_create(sPAPRMachineState *spapr)
1231 {
1232     DeviceState *dev = qdev_create(NULL, TYPE_SPAPR_RTC);
1233 
1234     qdev_init_nofail(dev);
1235     spapr->rtc = dev;
1236 
1237     object_property_add_alias(qdev_get_machine(), "rtc-time",
1238                               OBJECT(spapr->rtc), "date", NULL);
1239 }
1240 
1241 /* Returns whether we want to use VGA or not */
1242 static int spapr_vga_init(PCIBus *pci_bus)
1243 {
1244     switch (vga_interface_type) {
1245     case VGA_NONE:
1246         return false;
1247     case VGA_DEVICE:
1248         return true;
1249     case VGA_STD:
1250     case VGA_VIRTIO:
1251         return pci_vga_init(pci_bus) != NULL;
1252     default:
1253         fprintf(stderr, "This vga model is not supported,"
1254                 "currently it only supports -vga std\n");
1255         exit(0);
1256     }
1257 }
1258 
1259 static int spapr_post_load(void *opaque, int version_id)
1260 {
1261     sPAPRMachineState *spapr = (sPAPRMachineState *)opaque;
1262     int err = 0;
1263 
1264     /* In earlier versions, there was no separate qdev for the PAPR
1265      * RTC, so the RTC offset was stored directly in sPAPREnvironment.
1266      * So when migrating from those versions, poke the incoming offset
1267      * value into the RTC device */
1268     if (version_id < 3) {
1269         err = spapr_rtc_import_offset(spapr->rtc, spapr->rtc_offset);
1270     }
1271 
1272     return err;
1273 }
1274 
1275 static bool version_before_3(void *opaque, int version_id)
1276 {
1277     return version_id < 3;
1278 }
1279 
1280 static const VMStateDescription vmstate_spapr = {
1281     .name = "spapr",
1282     .version_id = 3,
1283     .minimum_version_id = 1,
1284     .post_load = spapr_post_load,
1285     .fields = (VMStateField[]) {
1286         /* used to be @next_irq */
1287         VMSTATE_UNUSED_BUFFER(version_before_3, 0, 4),
1288 
1289         /* RTC offset */
1290         VMSTATE_UINT64_TEST(rtc_offset, sPAPRMachineState, version_before_3),
1291 
1292         VMSTATE_PPC_TIMEBASE_V(tb, sPAPRMachineState, 2),
1293         VMSTATE_END_OF_LIST()
1294     },
1295 };
1296 
1297 static int htab_save_setup(QEMUFile *f, void *opaque)
1298 {
1299     sPAPRMachineState *spapr = opaque;
1300 
1301     /* "Iteration" header */
1302     qemu_put_be32(f, spapr->htab_shift);
1303 
1304     if (spapr->htab) {
1305         spapr->htab_save_index = 0;
1306         spapr->htab_first_pass = true;
1307     } else {
1308         assert(kvm_enabled());
1309 
1310         spapr->htab_fd = kvmppc_get_htab_fd(false);
1311         spapr->htab_fd_stale = false;
1312         if (spapr->htab_fd < 0) {
1313             fprintf(stderr, "Unable to open fd for reading hash table from KVM: %s\n",
1314                     strerror(errno));
1315             return -1;
1316         }
1317     }
1318 
1319 
1320     return 0;
1321 }
1322 
1323 static void htab_save_first_pass(QEMUFile *f, sPAPRMachineState *spapr,
1324                                  int64_t max_ns)
1325 {
1326     int htabslots = HTAB_SIZE(spapr) / HASH_PTE_SIZE_64;
1327     int index = spapr->htab_save_index;
1328     int64_t starttime = qemu_clock_get_ns(QEMU_CLOCK_REALTIME);
1329 
1330     assert(spapr->htab_first_pass);
1331 
1332     do {
1333         int chunkstart;
1334 
1335         /* Consume invalid HPTEs */
1336         while ((index < htabslots)
1337                && !HPTE_VALID(HPTE(spapr->htab, index))) {
1338             index++;
1339             CLEAN_HPTE(HPTE(spapr->htab, index));
1340         }
1341 
1342         /* Consume valid HPTEs */
1343         chunkstart = index;
1344         while ((index < htabslots) && (index - chunkstart < USHRT_MAX)
1345                && HPTE_VALID(HPTE(spapr->htab, index))) {
1346             index++;
1347             CLEAN_HPTE(HPTE(spapr->htab, index));
1348         }
1349 
1350         if (index > chunkstart) {
1351             int n_valid = index - chunkstart;
1352 
1353             qemu_put_be32(f, chunkstart);
1354             qemu_put_be16(f, n_valid);
1355             qemu_put_be16(f, 0);
1356             qemu_put_buffer(f, HPTE(spapr->htab, chunkstart),
1357                             HASH_PTE_SIZE_64 * n_valid);
1358 
1359             if ((qemu_clock_get_ns(QEMU_CLOCK_REALTIME) - starttime) > max_ns) {
1360                 break;
1361             }
1362         }
1363     } while ((index < htabslots) && !qemu_file_rate_limit(f));
1364 
1365     if (index >= htabslots) {
1366         assert(index == htabslots);
1367         index = 0;
1368         spapr->htab_first_pass = false;
1369     }
1370     spapr->htab_save_index = index;
1371 }
1372 
1373 static int htab_save_later_pass(QEMUFile *f, sPAPRMachineState *spapr,
1374                                 int64_t max_ns)
1375 {
1376     bool final = max_ns < 0;
1377     int htabslots = HTAB_SIZE(spapr) / HASH_PTE_SIZE_64;
1378     int examined = 0, sent = 0;
1379     int index = spapr->htab_save_index;
1380     int64_t starttime = qemu_clock_get_ns(QEMU_CLOCK_REALTIME);
1381 
1382     assert(!spapr->htab_first_pass);
1383 
1384     do {
1385         int chunkstart, invalidstart;
1386 
1387         /* Consume non-dirty HPTEs */
1388         while ((index < htabslots)
1389                && !HPTE_DIRTY(HPTE(spapr->htab, index))) {
1390             index++;
1391             examined++;
1392         }
1393 
1394         chunkstart = index;
1395         /* Consume valid dirty HPTEs */
1396         while ((index < htabslots) && (index - chunkstart < USHRT_MAX)
1397                && HPTE_DIRTY(HPTE(spapr->htab, index))
1398                && HPTE_VALID(HPTE(spapr->htab, index))) {
1399             CLEAN_HPTE(HPTE(spapr->htab, index));
1400             index++;
1401             examined++;
1402         }
1403 
1404         invalidstart = index;
1405         /* Consume invalid dirty HPTEs */
1406         while ((index < htabslots) && (index - invalidstart < USHRT_MAX)
1407                && HPTE_DIRTY(HPTE(spapr->htab, index))
1408                && !HPTE_VALID(HPTE(spapr->htab, index))) {
1409             CLEAN_HPTE(HPTE(spapr->htab, index));
1410             index++;
1411             examined++;
1412         }
1413 
1414         if (index > chunkstart) {
1415             int n_valid = invalidstart - chunkstart;
1416             int n_invalid = index - invalidstart;
1417 
1418             qemu_put_be32(f, chunkstart);
1419             qemu_put_be16(f, n_valid);
1420             qemu_put_be16(f, n_invalid);
1421             qemu_put_buffer(f, HPTE(spapr->htab, chunkstart),
1422                             HASH_PTE_SIZE_64 * n_valid);
1423             sent += index - chunkstart;
1424 
1425             if (!final && (qemu_clock_get_ns(QEMU_CLOCK_REALTIME) - starttime) > max_ns) {
1426                 break;
1427             }
1428         }
1429 
1430         if (examined >= htabslots) {
1431             break;
1432         }
1433 
1434         if (index >= htabslots) {
1435             assert(index == htabslots);
1436             index = 0;
1437         }
1438     } while ((examined < htabslots) && (!qemu_file_rate_limit(f) || final));
1439 
1440     if (index >= htabslots) {
1441         assert(index == htabslots);
1442         index = 0;
1443     }
1444 
1445     spapr->htab_save_index = index;
1446 
1447     return (examined >= htabslots) && (sent == 0) ? 1 : 0;
1448 }
1449 
1450 #define MAX_ITERATION_NS    5000000 /* 5 ms */
1451 #define MAX_KVM_BUF_SIZE    2048
1452 
1453 static int htab_save_iterate(QEMUFile *f, void *opaque)
1454 {
1455     sPAPRMachineState *spapr = opaque;
1456     int rc = 0;
1457 
1458     /* Iteration header */
1459     qemu_put_be32(f, 0);
1460 
1461     if (!spapr->htab) {
1462         assert(kvm_enabled());
1463 
1464         rc = spapr_check_htab_fd(spapr);
1465         if (rc < 0) {
1466             return rc;
1467         }
1468 
1469         rc = kvmppc_save_htab(f, spapr->htab_fd,
1470                               MAX_KVM_BUF_SIZE, MAX_ITERATION_NS);
1471         if (rc < 0) {
1472             return rc;
1473         }
1474     } else  if (spapr->htab_first_pass) {
1475         htab_save_first_pass(f, spapr, MAX_ITERATION_NS);
1476     } else {
1477         rc = htab_save_later_pass(f, spapr, MAX_ITERATION_NS);
1478     }
1479 
1480     /* End marker */
1481     qemu_put_be32(f, 0);
1482     qemu_put_be16(f, 0);
1483     qemu_put_be16(f, 0);
1484 
1485     return rc;
1486 }
1487 
1488 static int htab_save_complete(QEMUFile *f, void *opaque)
1489 {
1490     sPAPRMachineState *spapr = opaque;
1491 
1492     /* Iteration header */
1493     qemu_put_be32(f, 0);
1494 
1495     if (!spapr->htab) {
1496         int rc;
1497 
1498         assert(kvm_enabled());
1499 
1500         rc = spapr_check_htab_fd(spapr);
1501         if (rc < 0) {
1502             return rc;
1503         }
1504 
1505         rc = kvmppc_save_htab(f, spapr->htab_fd, MAX_KVM_BUF_SIZE, -1);
1506         if (rc < 0) {
1507             return rc;
1508         }
1509         close(spapr->htab_fd);
1510         spapr->htab_fd = -1;
1511     } else {
1512         htab_save_later_pass(f, spapr, -1);
1513     }
1514 
1515     /* End marker */
1516     qemu_put_be32(f, 0);
1517     qemu_put_be16(f, 0);
1518     qemu_put_be16(f, 0);
1519 
1520     return 0;
1521 }
1522 
1523 static int htab_load(QEMUFile *f, void *opaque, int version_id)
1524 {
1525     sPAPRMachineState *spapr = opaque;
1526     uint32_t section_hdr;
1527     int fd = -1;
1528 
1529     if (version_id < 1 || version_id > 1) {
1530         fprintf(stderr, "htab_load() bad version\n");
1531         return -EINVAL;
1532     }
1533 
1534     section_hdr = qemu_get_be32(f);
1535 
1536     if (section_hdr) {
1537         /* First section, just the hash shift */
1538         if (spapr->htab_shift != section_hdr) {
1539             error_report("htab_shift mismatch: source %d target %d",
1540                          section_hdr, spapr->htab_shift);
1541             return -EINVAL;
1542         }
1543         return 0;
1544     }
1545 
1546     if (!spapr->htab) {
1547         assert(kvm_enabled());
1548 
1549         fd = kvmppc_get_htab_fd(true);
1550         if (fd < 0) {
1551             fprintf(stderr, "Unable to open fd to restore KVM hash table: %s\n",
1552                     strerror(errno));
1553         }
1554     }
1555 
1556     while (true) {
1557         uint32_t index;
1558         uint16_t n_valid, n_invalid;
1559 
1560         index = qemu_get_be32(f);
1561         n_valid = qemu_get_be16(f);
1562         n_invalid = qemu_get_be16(f);
1563 
1564         if ((index == 0) && (n_valid == 0) && (n_invalid == 0)) {
1565             /* End of Stream */
1566             break;
1567         }
1568 
1569         if ((index + n_valid + n_invalid) >
1570             (HTAB_SIZE(spapr) / HASH_PTE_SIZE_64)) {
1571             /* Bad index in stream */
1572             fprintf(stderr, "htab_load() bad index %d (%hd+%hd entries) "
1573                     "in htab stream (htab_shift=%d)\n", index, n_valid, n_invalid,
1574                     spapr->htab_shift);
1575             return -EINVAL;
1576         }
1577 
1578         if (spapr->htab) {
1579             if (n_valid) {
1580                 qemu_get_buffer(f, HPTE(spapr->htab, index),
1581                                 HASH_PTE_SIZE_64 * n_valid);
1582             }
1583             if (n_invalid) {
1584                 memset(HPTE(spapr->htab, index + n_valid), 0,
1585                        HASH_PTE_SIZE_64 * n_invalid);
1586             }
1587         } else {
1588             int rc;
1589 
1590             assert(fd >= 0);
1591 
1592             rc = kvmppc_load_htab_chunk(f, fd, index, n_valid, n_invalid);
1593             if (rc < 0) {
1594                 return rc;
1595             }
1596         }
1597     }
1598 
1599     if (!spapr->htab) {
1600         assert(fd >= 0);
1601         close(fd);
1602     }
1603 
1604     return 0;
1605 }
1606 
1607 static SaveVMHandlers savevm_htab_handlers = {
1608     .save_live_setup = htab_save_setup,
1609     .save_live_iterate = htab_save_iterate,
1610     .save_live_complete_precopy = htab_save_complete,
1611     .load_state = htab_load,
1612 };
1613 
1614 static void spapr_boot_set(void *opaque, const char *boot_device,
1615                            Error **errp)
1616 {
1617     MachineState *machine = MACHINE(qdev_get_machine());
1618     machine->boot_order = g_strdup(boot_device);
1619 }
1620 
1621 static void spapr_cpu_init(sPAPRMachineState *spapr, PowerPCCPU *cpu)
1622 {
1623     CPUPPCState *env = &cpu->env;
1624 
1625     /* Set time-base frequency to 512 MHz */
1626     cpu_ppc_tb_init(env, TIMEBASE_FREQ);
1627 
1628     /* PAPR always has exception vectors in RAM not ROM. To ensure this,
1629      * MSR[IP] should never be set.
1630      */
1631     env->msr_mask &= ~(1 << 6);
1632 
1633     /* Tell KVM that we're in PAPR mode */
1634     if (kvm_enabled()) {
1635         kvmppc_set_papr(cpu);
1636     }
1637 
1638     if (cpu->max_compat) {
1639         if (ppc_set_compat(cpu, cpu->max_compat) < 0) {
1640             exit(1);
1641         }
1642     }
1643 
1644     xics_cpu_setup(spapr->icp, cpu);
1645 
1646     qemu_register_reset(spapr_cpu_reset, cpu);
1647 }
1648 
1649 /*
1650  * Reset routine for LMB DR devices.
1651  *
1652  * Unlike PCI DR devices, LMB DR devices explicitly register this reset
1653  * routine. Reset for PCI DR devices will be handled by PHB reset routine
1654  * when it walks all its children devices. LMB devices reset occurs
1655  * as part of spapr_ppc_reset().
1656  */
1657 static void spapr_drc_reset(void *opaque)
1658 {
1659     sPAPRDRConnector *drc = opaque;
1660     DeviceState *d = DEVICE(drc);
1661 
1662     if (d) {
1663         device_reset(d);
1664     }
1665 }
1666 
1667 static void spapr_create_lmb_dr_connectors(sPAPRMachineState *spapr)
1668 {
1669     MachineState *machine = MACHINE(spapr);
1670     uint64_t lmb_size = SPAPR_MEMORY_BLOCK_SIZE;
1671     uint32_t nr_lmbs = (machine->maxram_size - machine->ram_size)/lmb_size;
1672     int i;
1673 
1674     for (i = 0; i < nr_lmbs; i++) {
1675         sPAPRDRConnector *drc;
1676         uint64_t addr;
1677 
1678         addr = i * lmb_size + spapr->hotplug_memory.base;
1679         drc = spapr_dr_connector_new(OBJECT(spapr), SPAPR_DR_CONNECTOR_TYPE_LMB,
1680                                      addr/lmb_size);
1681         qemu_register_reset(spapr_drc_reset, drc);
1682     }
1683 }
1684 
1685 /*
1686  * If RAM size, maxmem size and individual node mem sizes aren't aligned
1687  * to SPAPR_MEMORY_BLOCK_SIZE(256MB), then refuse to start the guest
1688  * since we can't support such unaligned sizes with DRCONF_MEMORY.
1689  */
1690 static void spapr_validate_node_memory(MachineState *machine)
1691 {
1692     int i;
1693 
1694     if (machine->maxram_size % SPAPR_MEMORY_BLOCK_SIZE ||
1695         machine->ram_size % SPAPR_MEMORY_BLOCK_SIZE) {
1696         error_report("Can't support memory configuration where RAM size "
1697                      "0x" RAM_ADDR_FMT " or maxmem size "
1698                      "0x" RAM_ADDR_FMT " isn't aligned to %llu MB",
1699                      machine->ram_size, machine->maxram_size,
1700                      SPAPR_MEMORY_BLOCK_SIZE/M_BYTE);
1701         exit(EXIT_FAILURE);
1702     }
1703 
1704     for (i = 0; i < nb_numa_nodes; i++) {
1705         if (numa_info[i].node_mem % SPAPR_MEMORY_BLOCK_SIZE) {
1706             error_report("Can't support memory configuration where memory size"
1707                          " %" PRIx64 " of node %d isn't aligned to %llu MB",
1708                          numa_info[i].node_mem, i,
1709                          SPAPR_MEMORY_BLOCK_SIZE/M_BYTE);
1710             exit(EXIT_FAILURE);
1711         }
1712     }
1713 }
1714 
1715 /* pSeries LPAR / sPAPR hardware init */
1716 static void ppc_spapr_init(MachineState *machine)
1717 {
1718     sPAPRMachineState *spapr = SPAPR_MACHINE(machine);
1719     sPAPRMachineClass *smc = SPAPR_MACHINE_GET_CLASS(machine);
1720     const char *kernel_filename = machine->kernel_filename;
1721     const char *kernel_cmdline = machine->kernel_cmdline;
1722     const char *initrd_filename = machine->initrd_filename;
1723     PowerPCCPU *cpu;
1724     PCIHostState *phb;
1725     int i;
1726     MemoryRegion *sysmem = get_system_memory();
1727     MemoryRegion *ram = g_new(MemoryRegion, 1);
1728     MemoryRegion *rma_region;
1729     void *rma = NULL;
1730     hwaddr rma_alloc_size;
1731     hwaddr node0_size = spapr_node0_size();
1732     uint32_t initrd_base = 0;
1733     long kernel_size = 0, initrd_size = 0;
1734     long load_limit, fw_size;
1735     bool kernel_le = false;
1736     char *filename;
1737 
1738     msi_supported = true;
1739 
1740     QLIST_INIT(&spapr->phbs);
1741 
1742     cpu_ppc_hypercall = emulate_spapr_hypercall;
1743 
1744     /* Allocate RMA if necessary */
1745     rma_alloc_size = kvmppc_alloc_rma(&rma);
1746 
1747     if (rma_alloc_size == -1) {
1748         error_report("Unable to create RMA");
1749         exit(1);
1750     }
1751 
1752     if (rma_alloc_size && (rma_alloc_size < node0_size)) {
1753         spapr->rma_size = rma_alloc_size;
1754     } else {
1755         spapr->rma_size = node0_size;
1756 
1757         /* With KVM, we don't actually know whether KVM supports an
1758          * unbounded RMA (PR KVM) or is limited by the hash table size
1759          * (HV KVM using VRMA), so we always assume the latter
1760          *
1761          * In that case, we also limit the initial allocations for RTAS
1762          * etc... to 256M since we have no way to know what the VRMA size
1763          * is going to be as it depends on the size of the hash table
1764          * isn't determined yet.
1765          */
1766         if (kvm_enabled()) {
1767             spapr->vrma_adjust = 1;
1768             spapr->rma_size = MIN(spapr->rma_size, 0x10000000);
1769         }
1770     }
1771 
1772     if (spapr->rma_size > node0_size) {
1773         fprintf(stderr, "Error: Numa node 0 has to span the RMA (%#08"HWADDR_PRIx")\n",
1774                 spapr->rma_size);
1775         exit(1);
1776     }
1777 
1778     /* Setup a load limit for the ramdisk leaving room for SLOF and FDT */
1779     load_limit = MIN(spapr->rma_size, RTAS_MAX_ADDR) - FW_OVERHEAD;
1780 
1781     /* We aim for a hash table of size 1/128 the size of RAM.  The
1782      * normal rule of thumb is 1/64 the size of RAM, but that's much
1783      * more than needed for the Linux guests we support. */
1784     spapr->htab_shift = 18; /* Minimum architected size */
1785     while (spapr->htab_shift <= 46) {
1786         if ((1ULL << (spapr->htab_shift + 7)) >= machine->maxram_size) {
1787             break;
1788         }
1789         spapr->htab_shift++;
1790     }
1791     spapr_alloc_htab(spapr);
1792 
1793     /* Set up Interrupt Controller before we create the VCPUs */
1794     spapr->icp = xics_system_init(machine,
1795                                   DIV_ROUND_UP(max_cpus * kvmppc_smt_threads(),
1796                                                smp_threads),
1797                                   XICS_IRQS);
1798 
1799     if (smc->dr_lmb_enabled) {
1800         spapr_validate_node_memory(machine);
1801     }
1802 
1803     /* init CPUs */
1804     if (machine->cpu_model == NULL) {
1805         machine->cpu_model = kvm_enabled() ? "host" : "POWER7";
1806     }
1807     for (i = 0; i < smp_cpus; i++) {
1808         cpu = cpu_ppc_init(machine->cpu_model);
1809         if (cpu == NULL) {
1810             fprintf(stderr, "Unable to find PowerPC CPU definition\n");
1811             exit(1);
1812         }
1813         spapr_cpu_init(spapr, cpu);
1814     }
1815 
1816     if (kvm_enabled()) {
1817         /* Enable H_LOGICAL_CI_* so SLOF can talk to in-kernel devices */
1818         kvmppc_enable_logical_ci_hcalls();
1819         kvmppc_enable_set_mode_hcall();
1820     }
1821 
1822     /* allocate RAM */
1823     memory_region_allocate_system_memory(ram, NULL, "ppc_spapr.ram",
1824                                          machine->ram_size);
1825     memory_region_add_subregion(sysmem, 0, ram);
1826 
1827     if (rma_alloc_size && rma) {
1828         rma_region = g_new(MemoryRegion, 1);
1829         memory_region_init_ram_ptr(rma_region, NULL, "ppc_spapr.rma",
1830                                    rma_alloc_size, rma);
1831         vmstate_register_ram_global(rma_region);
1832         memory_region_add_subregion(sysmem, 0, rma_region);
1833     }
1834 
1835     /* initialize hotplug memory address space */
1836     if (machine->ram_size < machine->maxram_size) {
1837         ram_addr_t hotplug_mem_size = machine->maxram_size - machine->ram_size;
1838 
1839         if (machine->ram_slots > SPAPR_MAX_RAM_SLOTS) {
1840             error_report("Specified number of memory slots %" PRIu64
1841                          " exceeds max supported %d",
1842                          machine->ram_slots, SPAPR_MAX_RAM_SLOTS);
1843             exit(EXIT_FAILURE);
1844         }
1845 
1846         spapr->hotplug_memory.base = ROUND_UP(machine->ram_size,
1847                                               SPAPR_HOTPLUG_MEM_ALIGN);
1848         memory_region_init(&spapr->hotplug_memory.mr, OBJECT(spapr),
1849                            "hotplug-memory", hotplug_mem_size);
1850         memory_region_add_subregion(sysmem, spapr->hotplug_memory.base,
1851                                     &spapr->hotplug_memory.mr);
1852     }
1853 
1854     if (smc->dr_lmb_enabled) {
1855         spapr_create_lmb_dr_connectors(spapr);
1856     }
1857 
1858     filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, "spapr-rtas.bin");
1859     if (!filename) {
1860         error_report("Could not find LPAR rtas '%s'", "spapr-rtas.bin");
1861         exit(1);
1862     }
1863     spapr->rtas_size = get_image_size(filename);
1864     spapr->rtas_blob = g_malloc(spapr->rtas_size);
1865     if (load_image_size(filename, spapr->rtas_blob, spapr->rtas_size) < 0) {
1866         error_report("Could not load LPAR rtas '%s'", filename);
1867         exit(1);
1868     }
1869     if (spapr->rtas_size > RTAS_MAX_SIZE) {
1870         error_report("RTAS too big ! 0x%zx bytes (max is 0x%x)",
1871                      (size_t)spapr->rtas_size, RTAS_MAX_SIZE);
1872         exit(1);
1873     }
1874     g_free(filename);
1875 
1876     /* Set up EPOW events infrastructure */
1877     spapr_events_init(spapr);
1878 
1879     /* Set up the RTC RTAS interfaces */
1880     spapr_rtc_create(spapr);
1881 
1882     /* Set up VIO bus */
1883     spapr->vio_bus = spapr_vio_bus_init();
1884 
1885     for (i = 0; i < MAX_SERIAL_PORTS; i++) {
1886         if (serial_hds[i]) {
1887             spapr_vty_create(spapr->vio_bus, serial_hds[i]);
1888         }
1889     }
1890 
1891     /* We always have at least the nvram device on VIO */
1892     spapr_create_nvram(spapr);
1893 
1894     /* Set up PCI */
1895     spapr_pci_rtas_init();
1896 
1897     phb = spapr_create_phb(spapr, 0);
1898 
1899     for (i = 0; i < nb_nics; i++) {
1900         NICInfo *nd = &nd_table[i];
1901 
1902         if (!nd->model) {
1903             nd->model = g_strdup("ibmveth");
1904         }
1905 
1906         if (strcmp(nd->model, "ibmveth") == 0) {
1907             spapr_vlan_create(spapr->vio_bus, nd);
1908         } else {
1909             pci_nic_init_nofail(&nd_table[i], phb->bus, nd->model, NULL);
1910         }
1911     }
1912 
1913     for (i = 0; i <= drive_get_max_bus(IF_SCSI); i++) {
1914         spapr_vscsi_create(spapr->vio_bus);
1915     }
1916 
1917     /* Graphics */
1918     if (spapr_vga_init(phb->bus)) {
1919         spapr->has_graphics = true;
1920         machine->usb |= defaults_enabled() && !machine->usb_disabled;
1921     }
1922 
1923     if (machine->usb) {
1924         if (smc->use_ohci_by_default) {
1925             pci_create_simple(phb->bus, -1, "pci-ohci");
1926         } else {
1927             pci_create_simple(phb->bus, -1, "nec-usb-xhci");
1928         }
1929 
1930         if (spapr->has_graphics) {
1931             USBBus *usb_bus = usb_bus_find(-1);
1932 
1933             usb_create_simple(usb_bus, "usb-kbd");
1934             usb_create_simple(usb_bus, "usb-mouse");
1935         }
1936     }
1937 
1938     if (spapr->rma_size < (MIN_RMA_SLOF << 20)) {
1939         fprintf(stderr, "qemu: pSeries SLOF firmware requires >= "
1940                 "%ldM guest RMA (Real Mode Area memory)\n", MIN_RMA_SLOF);
1941         exit(1);
1942     }
1943 
1944     if (kernel_filename) {
1945         uint64_t lowaddr = 0;
1946 
1947         kernel_size = load_elf(kernel_filename, translate_kernel_address, NULL,
1948                                NULL, &lowaddr, NULL, 1, PPC_ELF_MACHINE, 0);
1949         if (kernel_size == ELF_LOAD_WRONG_ENDIAN) {
1950             kernel_size = load_elf(kernel_filename,
1951                                    translate_kernel_address, NULL,
1952                                    NULL, &lowaddr, NULL, 0, PPC_ELF_MACHINE, 0);
1953             kernel_le = kernel_size > 0;
1954         }
1955         if (kernel_size < 0) {
1956             fprintf(stderr, "qemu: error loading %s: %s\n",
1957                     kernel_filename, load_elf_strerror(kernel_size));
1958             exit(1);
1959         }
1960 
1961         /* load initrd */
1962         if (initrd_filename) {
1963             /* Try to locate the initrd in the gap between the kernel
1964              * and the firmware. Add a bit of space just in case
1965              */
1966             initrd_base = (KERNEL_LOAD_ADDR + kernel_size + 0x1ffff) & ~0xffff;
1967             initrd_size = load_image_targphys(initrd_filename, initrd_base,
1968                                               load_limit - initrd_base);
1969             if (initrd_size < 0) {
1970                 fprintf(stderr, "qemu: could not load initial ram disk '%s'\n",
1971                         initrd_filename);
1972                 exit(1);
1973             }
1974         } else {
1975             initrd_base = 0;
1976             initrd_size = 0;
1977         }
1978     }
1979 
1980     if (bios_name == NULL) {
1981         bios_name = FW_FILE_NAME;
1982     }
1983     filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, bios_name);
1984     if (!filename) {
1985         error_report("Could not find LPAR firmware '%s'", bios_name);
1986         exit(1);
1987     }
1988     fw_size = load_image_targphys(filename, 0, FW_MAX_SIZE);
1989     if (fw_size <= 0) {
1990         error_report("Could not load LPAR firmware '%s'", filename);
1991         exit(1);
1992     }
1993     g_free(filename);
1994 
1995     /* FIXME: Should register things through the MachineState's qdev
1996      * interface, this is a legacy from the sPAPREnvironment structure
1997      * which predated MachineState but had a similar function */
1998     vmstate_register(NULL, 0, &vmstate_spapr, spapr);
1999     register_savevm_live(NULL, "spapr/htab", -1, 1,
2000                          &savevm_htab_handlers, spapr);
2001 
2002     /* Prepare the device tree */
2003     spapr->fdt_skel = spapr_create_fdt_skel(initrd_base, initrd_size,
2004                                             kernel_size, kernel_le,
2005                                             kernel_cmdline,
2006                                             spapr->check_exception_irq);
2007     assert(spapr->fdt_skel != NULL);
2008 
2009     /* used by RTAS */
2010     QTAILQ_INIT(&spapr->ccs_list);
2011     qemu_register_reset(spapr_ccs_reset_hook, spapr);
2012 
2013     qemu_register_boot_set(spapr_boot_set, spapr);
2014 }
2015 
2016 static int spapr_kvm_type(const char *vm_type)
2017 {
2018     if (!vm_type) {
2019         return 0;
2020     }
2021 
2022     if (!strcmp(vm_type, "HV")) {
2023         return 1;
2024     }
2025 
2026     if (!strcmp(vm_type, "PR")) {
2027         return 2;
2028     }
2029 
2030     error_report("Unknown kvm-type specified '%s'", vm_type);
2031     exit(1);
2032 }
2033 
2034 /*
2035  * Implementation of an interface to adjust firmware path
2036  * for the bootindex property handling.
2037  */
2038 static char *spapr_get_fw_dev_path(FWPathProvider *p, BusState *bus,
2039                                    DeviceState *dev)
2040 {
2041 #define CAST(type, obj, name) \
2042     ((type *)object_dynamic_cast(OBJECT(obj), (name)))
2043     SCSIDevice *d = CAST(SCSIDevice,  dev, TYPE_SCSI_DEVICE);
2044     sPAPRPHBState *phb = CAST(sPAPRPHBState, dev, TYPE_SPAPR_PCI_HOST_BRIDGE);
2045 
2046     if (d) {
2047         void *spapr = CAST(void, bus->parent, "spapr-vscsi");
2048         VirtIOSCSI *virtio = CAST(VirtIOSCSI, bus->parent, TYPE_VIRTIO_SCSI);
2049         USBDevice *usb = CAST(USBDevice, bus->parent, TYPE_USB_DEVICE);
2050 
2051         if (spapr) {
2052             /*
2053              * Replace "channel@0/disk@0,0" with "disk@8000000000000000":
2054              * We use SRP luns of the form 8000 | (bus << 8) | (id << 5) | lun
2055              * in the top 16 bits of the 64-bit LUN
2056              */
2057             unsigned id = 0x8000 | (d->id << 8) | d->lun;
2058             return g_strdup_printf("%s@%"PRIX64, qdev_fw_name(dev),
2059                                    (uint64_t)id << 48);
2060         } else if (virtio) {
2061             /*
2062              * We use SRP luns of the form 01000000 | (target << 8) | lun
2063              * in the top 32 bits of the 64-bit LUN
2064              * Note: the quote above is from SLOF and it is wrong,
2065              * the actual binding is:
2066              * swap 0100 or 10 << or 20 << ( target lun-id -- srplun )
2067              */
2068             unsigned id = 0x1000000 | (d->id << 16) | d->lun;
2069             return g_strdup_printf("%s@%"PRIX64, qdev_fw_name(dev),
2070                                    (uint64_t)id << 32);
2071         } else if (usb) {
2072             /*
2073              * We use SRP luns of the form 01000000 | (usb-port << 16) | lun
2074              * in the top 32 bits of the 64-bit LUN
2075              */
2076             unsigned usb_port = atoi(usb->port->path);
2077             unsigned id = 0x1000000 | (usb_port << 16) | d->lun;
2078             return g_strdup_printf("%s@%"PRIX64, qdev_fw_name(dev),
2079                                    (uint64_t)id << 32);
2080         }
2081     }
2082 
2083     if (phb) {
2084         /* Replace "pci" with "pci@800000020000000" */
2085         return g_strdup_printf("pci@%"PRIX64, phb->buid);
2086     }
2087 
2088     return NULL;
2089 }
2090 
2091 static char *spapr_get_kvm_type(Object *obj, Error **errp)
2092 {
2093     sPAPRMachineState *spapr = SPAPR_MACHINE(obj);
2094 
2095     return g_strdup(spapr->kvm_type);
2096 }
2097 
2098 static void spapr_set_kvm_type(Object *obj, const char *value, Error **errp)
2099 {
2100     sPAPRMachineState *spapr = SPAPR_MACHINE(obj);
2101 
2102     g_free(spapr->kvm_type);
2103     spapr->kvm_type = g_strdup(value);
2104 }
2105 
2106 static void spapr_machine_initfn(Object *obj)
2107 {
2108     object_property_add_str(obj, "kvm-type",
2109                             spapr_get_kvm_type, spapr_set_kvm_type, NULL);
2110     object_property_set_description(obj, "kvm-type",
2111                                     "Specifies the KVM virtualization mode (HV, PR)",
2112                                     NULL);
2113 }
2114 
2115 static void spapr_machine_finalizefn(Object *obj)
2116 {
2117     sPAPRMachineState *spapr = SPAPR_MACHINE(obj);
2118 
2119     g_free(spapr->kvm_type);
2120 }
2121 
2122 static void ppc_cpu_do_nmi_on_cpu(void *arg)
2123 {
2124     CPUState *cs = arg;
2125 
2126     cpu_synchronize_state(cs);
2127     ppc_cpu_do_system_reset(cs);
2128 }
2129 
2130 static void spapr_nmi(NMIState *n, int cpu_index, Error **errp)
2131 {
2132     CPUState *cs;
2133 
2134     CPU_FOREACH(cs) {
2135         async_run_on_cpu(cs, ppc_cpu_do_nmi_on_cpu, cs);
2136     }
2137 }
2138 
2139 static void spapr_add_lmbs(DeviceState *dev, uint64_t addr, uint64_t size,
2140                            uint32_t node, Error **errp)
2141 {
2142     sPAPRDRConnector *drc;
2143     sPAPRDRConnectorClass *drck;
2144     uint32_t nr_lmbs = size/SPAPR_MEMORY_BLOCK_SIZE;
2145     int i, fdt_offset, fdt_size;
2146     void *fdt;
2147 
2148     /*
2149      * Check for DRC connectors and send hotplug notification to the
2150      * guest only in case of hotplugged memory. This allows cold plugged
2151      * memory to be specified at boot time.
2152      */
2153     if (!dev->hotplugged) {
2154         return;
2155     }
2156 
2157     for (i = 0; i < nr_lmbs; i++) {
2158         drc = spapr_dr_connector_by_id(SPAPR_DR_CONNECTOR_TYPE_LMB,
2159                 addr/SPAPR_MEMORY_BLOCK_SIZE);
2160         g_assert(drc);
2161 
2162         fdt = create_device_tree(&fdt_size);
2163         fdt_offset = spapr_populate_memory_node(fdt, node, addr,
2164                                                 SPAPR_MEMORY_BLOCK_SIZE);
2165 
2166         drck = SPAPR_DR_CONNECTOR_GET_CLASS(drc);
2167         drck->attach(drc, dev, fdt, fdt_offset, !dev->hotplugged, errp);
2168         addr += SPAPR_MEMORY_BLOCK_SIZE;
2169     }
2170     spapr_hotplug_req_add_by_count(SPAPR_DR_CONNECTOR_TYPE_LMB, nr_lmbs);
2171 }
2172 
2173 static void spapr_memory_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
2174                               uint32_t node, Error **errp)
2175 {
2176     Error *local_err = NULL;
2177     sPAPRMachineState *ms = SPAPR_MACHINE(hotplug_dev);
2178     PCDIMMDevice *dimm = PC_DIMM(dev);
2179     PCDIMMDeviceClass *ddc = PC_DIMM_GET_CLASS(dimm);
2180     MemoryRegion *mr = ddc->get_memory_region(dimm);
2181     uint64_t align = memory_region_get_alignment(mr);
2182     uint64_t size = memory_region_size(mr);
2183     uint64_t addr;
2184 
2185     if (size % SPAPR_MEMORY_BLOCK_SIZE) {
2186         error_setg(&local_err, "Hotplugged memory size must be a multiple of "
2187                       "%lld MB", SPAPR_MEMORY_BLOCK_SIZE/M_BYTE);
2188         goto out;
2189     }
2190 
2191     pc_dimm_memory_plug(dev, &ms->hotplug_memory, mr, align, &local_err);
2192     if (local_err) {
2193         goto out;
2194     }
2195 
2196     addr = object_property_get_int(OBJECT(dimm), PC_DIMM_ADDR_PROP, &local_err);
2197     if (local_err) {
2198         pc_dimm_memory_unplug(dev, &ms->hotplug_memory, mr);
2199         goto out;
2200     }
2201 
2202     spapr_add_lmbs(dev, addr, size, node, &error_abort);
2203 
2204 out:
2205     error_propagate(errp, local_err);
2206 }
2207 
2208 static void spapr_machine_device_plug(HotplugHandler *hotplug_dev,
2209                                       DeviceState *dev, Error **errp)
2210 {
2211     sPAPRMachineClass *smc = SPAPR_MACHINE_GET_CLASS(qdev_get_machine());
2212 
2213     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2214         int node;
2215 
2216         if (!smc->dr_lmb_enabled) {
2217             error_setg(errp, "Memory hotplug not supported for this machine");
2218             return;
2219         }
2220         node = object_property_get_int(OBJECT(dev), PC_DIMM_NODE_PROP, errp);
2221         if (*errp) {
2222             return;
2223         }
2224 
2225         /*
2226          * Currently PowerPC kernel doesn't allow hot-adding memory to
2227          * memory-less node, but instead will silently add the memory
2228          * to the first node that has some memory. This causes two
2229          * unexpected behaviours for the user.
2230          *
2231          * - Memory gets hotplugged to a different node than what the user
2232          *   specified.
2233          * - Since pc-dimm subsystem in QEMU still thinks that memory belongs
2234          *   to memory-less node, a reboot will set things accordingly
2235          *   and the previously hotplugged memory now ends in the right node.
2236          *   This appears as if some memory moved from one node to another.
2237          *
2238          * So until kernel starts supporting memory hotplug to memory-less
2239          * nodes, just prevent such attempts upfront in QEMU.
2240          */
2241         if (nb_numa_nodes && !numa_info[node].node_mem) {
2242             error_setg(errp, "Can't hotplug memory to memory-less node %d",
2243                        node);
2244             return;
2245         }
2246 
2247         spapr_memory_plug(hotplug_dev, dev, node, errp);
2248     }
2249 }
2250 
2251 static void spapr_machine_device_unplug(HotplugHandler *hotplug_dev,
2252                                       DeviceState *dev, Error **errp)
2253 {
2254     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2255         error_setg(errp, "Memory hot unplug not supported by sPAPR");
2256     }
2257 }
2258 
2259 static HotplugHandler *spapr_get_hotpug_handler(MachineState *machine,
2260                                              DeviceState *dev)
2261 {
2262     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2263         return HOTPLUG_HANDLER(machine);
2264     }
2265     return NULL;
2266 }
2267 
2268 static unsigned spapr_cpu_index_to_socket_id(unsigned cpu_index)
2269 {
2270     /* Allocate to NUMA nodes on a "socket" basis (not that concept of
2271      * socket means much for the paravirtualized PAPR platform) */
2272     return cpu_index / smp_threads / smp_cores;
2273 }
2274 
2275 static void spapr_machine_class_init(ObjectClass *oc, void *data)
2276 {
2277     MachineClass *mc = MACHINE_CLASS(oc);
2278     sPAPRMachineClass *smc = SPAPR_MACHINE_CLASS(oc);
2279     FWPathProviderClass *fwc = FW_PATH_PROVIDER_CLASS(oc);
2280     NMIClass *nc = NMI_CLASS(oc);
2281     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
2282 
2283     mc->desc = "pSeries Logical Partition (PAPR compliant)";
2284 
2285     /*
2286      * We set up the default / latest behaviour here.  The class_init
2287      * functions for the specific versioned machine types can override
2288      * these details for backwards compatibility
2289      */
2290     mc->init = ppc_spapr_init;
2291     mc->reset = ppc_spapr_reset;
2292     mc->block_default_type = IF_SCSI;
2293     mc->max_cpus = MAX_CPUMASK_BITS;
2294     mc->no_parallel = 1;
2295     mc->default_boot_order = "";
2296     mc->default_ram_size = 512 * M_BYTE;
2297     mc->kvm_type = spapr_kvm_type;
2298     mc->has_dynamic_sysbus = true;
2299     mc->pci_allow_0_address = true;
2300     mc->get_hotplug_handler = spapr_get_hotpug_handler;
2301     hc->plug = spapr_machine_device_plug;
2302     hc->unplug = spapr_machine_device_unplug;
2303     mc->cpu_index_to_socket_id = spapr_cpu_index_to_socket_id;
2304 
2305     smc->dr_lmb_enabled = true;
2306     fwc->get_dev_path = spapr_get_fw_dev_path;
2307     nc->nmi_monitor_handler = spapr_nmi;
2308 }
2309 
2310 static const TypeInfo spapr_machine_info = {
2311     .name          = TYPE_SPAPR_MACHINE,
2312     .parent        = TYPE_MACHINE,
2313     .abstract      = true,
2314     .instance_size = sizeof(sPAPRMachineState),
2315     .instance_init = spapr_machine_initfn,
2316     .instance_finalize = spapr_machine_finalizefn,
2317     .class_size    = sizeof(sPAPRMachineClass),
2318     .class_init    = spapr_machine_class_init,
2319     .interfaces = (InterfaceInfo[]) {
2320         { TYPE_FW_PATH_PROVIDER },
2321         { TYPE_NMI },
2322         { TYPE_HOTPLUG_HANDLER },
2323         { }
2324     },
2325 };
2326 
2327 #define DEFINE_SPAPR_MACHINE(suffix, verstr, latest)                 \
2328     static void spapr_machine_##suffix##_class_init(ObjectClass *oc, \
2329                                                     void *data)      \
2330     {                                                                \
2331         MachineClass *mc = MACHINE_CLASS(oc);                        \
2332         spapr_machine_##suffix##_class_options(mc);                  \
2333         if (latest) {                                                \
2334             mc->alias = "pseries";                                   \
2335             mc->is_default = 1;                                      \
2336         }                                                            \
2337     }                                                                \
2338     static void spapr_machine_##suffix##_instance_init(Object *obj)  \
2339     {                                                                \
2340         MachineState *machine = MACHINE(obj);                        \
2341         spapr_machine_##suffix##_instance_options(machine);          \
2342     }                                                                \
2343     static const TypeInfo spapr_machine_##suffix##_info = {          \
2344         .name = MACHINE_TYPE_NAME("pseries-" verstr),                \
2345         .parent = TYPE_SPAPR_MACHINE,                                \
2346         .class_init = spapr_machine_##suffix##_class_init,           \
2347         .instance_init = spapr_machine_##suffix##_instance_init,     \
2348     };                                                               \
2349     static void spapr_machine_register_##suffix(void)                \
2350     {                                                                \
2351         type_register(&spapr_machine_##suffix##_info);               \
2352     }                                                                \
2353     machine_init(spapr_machine_register_##suffix)
2354 
2355 /*
2356  * pseries-2.6
2357  */
2358 static void spapr_machine_2_6_instance_options(MachineState *machine)
2359 {
2360 }
2361 
2362 static void spapr_machine_2_6_class_options(MachineClass *mc)
2363 {
2364     /* Defaults for the latest behaviour inherited from the base class */
2365 }
2366 
2367 DEFINE_SPAPR_MACHINE(2_6, "2.6", true);
2368 
2369 /*
2370  * pseries-2.5
2371  */
2372 #define SPAPR_COMPAT_2_5 \
2373         HW_COMPAT_2_5
2374 
2375 static void spapr_machine_2_5_instance_options(MachineState *machine)
2376 {
2377 }
2378 
2379 static void spapr_machine_2_5_class_options(MachineClass *mc)
2380 {
2381     sPAPRMachineClass *smc = SPAPR_MACHINE_CLASS(mc);
2382 
2383     spapr_machine_2_6_class_options(mc);
2384     smc->use_ohci_by_default = true;
2385     SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_5);
2386 }
2387 
2388 DEFINE_SPAPR_MACHINE(2_5, "2.5", false);
2389 
2390 /*
2391  * pseries-2.4
2392  */
2393 #define SPAPR_COMPAT_2_4 \
2394         HW_COMPAT_2_4
2395 
2396 static void spapr_machine_2_4_instance_options(MachineState *machine)
2397 {
2398     spapr_machine_2_5_instance_options(machine);
2399 }
2400 
2401 static void spapr_machine_2_4_class_options(MachineClass *mc)
2402 {
2403     sPAPRMachineClass *smc = SPAPR_MACHINE_CLASS(mc);
2404 
2405     spapr_machine_2_5_class_options(mc);
2406     smc->dr_lmb_enabled = false;
2407     SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_4);
2408 }
2409 
2410 DEFINE_SPAPR_MACHINE(2_4, "2.4", false);
2411 
2412 /*
2413  * pseries-2.3
2414  */
2415 #define SPAPR_COMPAT_2_3 \
2416         SPAPR_COMPAT_2_4 \
2417         HW_COMPAT_2_3 \
2418         {\
2419             .driver   = "spapr-pci-host-bridge",\
2420             .property = "dynamic-reconfiguration",\
2421             .value    = "off",\
2422         },
2423 
2424 static void spapr_machine_2_3_instance_options(MachineState *machine)
2425 {
2426     spapr_machine_2_4_instance_options(machine);
2427     savevm_skip_section_footers();
2428     global_state_set_optional();
2429 }
2430 
2431 static void spapr_machine_2_3_class_options(MachineClass *mc)
2432 {
2433     spapr_machine_2_4_class_options(mc);
2434     SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_3);
2435 }
2436 DEFINE_SPAPR_MACHINE(2_3, "2.3", false);
2437 
2438 /*
2439  * pseries-2.2
2440  */
2441 
2442 #define SPAPR_COMPAT_2_2 \
2443         SPAPR_COMPAT_2_3 \
2444         HW_COMPAT_2_2 \
2445         {\
2446             .driver   = TYPE_SPAPR_PCI_HOST_BRIDGE,\
2447             .property = "mem_win_size",\
2448             .value    = "0x20000000",\
2449         },
2450 
2451 static void spapr_machine_2_2_instance_options(MachineState *machine)
2452 {
2453     spapr_machine_2_3_instance_options(machine);
2454 }
2455 
2456 static void spapr_machine_2_2_class_options(MachineClass *mc)
2457 {
2458     spapr_machine_2_3_class_options(mc);
2459     SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_2);
2460 }
2461 DEFINE_SPAPR_MACHINE(2_2, "2.2", false);
2462 
2463 /*
2464  * pseries-2.1
2465  */
2466 #define SPAPR_COMPAT_2_1 \
2467         SPAPR_COMPAT_2_2 \
2468         HW_COMPAT_2_1
2469 
2470 static void spapr_machine_2_1_instance_options(MachineState *machine)
2471 {
2472     spapr_machine_2_2_instance_options(machine);
2473 }
2474 
2475 static void spapr_machine_2_1_class_options(MachineClass *mc)
2476 {
2477     spapr_machine_2_2_class_options(mc);
2478     SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_1);
2479 }
2480 DEFINE_SPAPR_MACHINE(2_1, "2.1", false);
2481 
2482 static void spapr_machine_register_types(void)
2483 {
2484     type_register_static(&spapr_machine_info);
2485 }
2486 
2487 type_init(spapr_machine_register_types)
2488