xref: /openbmc/qemu/hw/ppc/spapr.c (revision 47dc0ec576c3d8748985e34d273c8dc0465c5ddb)
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 "qapi/error.h"
29 #include "sysemu/sysemu.h"
30 #include "sysemu/numa.h"
31 #include "hw/hw.h"
32 #include "qemu/log.h"
33 #include "hw/fw-path-provider.h"
34 #include "elf.h"
35 #include "net/net.h"
36 #include "sysemu/device_tree.h"
37 #include "sysemu/block-backend.h"
38 #include "sysemu/cpus.h"
39 #include "sysemu/kvm.h"
40 #include "sysemu/device_tree.h"
41 #include "kvm_ppc.h"
42 #include "migration/migration.h"
43 #include "mmu-hash64.h"
44 #include "qom/cpu.h"
45 
46 #include "hw/boards.h"
47 #include "hw/ppc/ppc.h"
48 #include "hw/loader.h"
49 
50 #include "hw/ppc/spapr.h"
51 #include "hw/ppc/spapr_vio.h"
52 #include "hw/pci-host/spapr.h"
53 #include "hw/ppc/xics.h"
54 #include "hw/pci/msi.h"
55 
56 #include "hw/pci/pci.h"
57 #include "hw/scsi/scsi.h"
58 #include "hw/virtio/virtio-scsi.h"
59 
60 #include "exec/address-spaces.h"
61 #include "hw/usb.h"
62 #include "qemu/config-file.h"
63 #include "qemu/error-report.h"
64 #include "trace.h"
65 #include "hw/nmi.h"
66 
67 #include "hw/compat.h"
68 #include "qemu/cutils.h"
69 #include "hw/ppc/spapr_cpu_core.h"
70 #include "qmp-commands.h"
71 
72 #include <libfdt.h>
73 
74 /* SLOF memory layout:
75  *
76  * SLOF raw image loaded at 0, copies its romfs right below the flat
77  * device-tree, then position SLOF itself 31M below that
78  *
79  * So we set FW_OVERHEAD to 40MB which should account for all of that
80  * and more
81  *
82  * We load our kernel at 4M, leaving space for SLOF initial image
83  */
84 #define FDT_MAX_SIZE            0x100000
85 #define RTAS_MAX_SIZE           0x10000
86 #define RTAS_MAX_ADDR           0x80000000 /* RTAS must stay below that */
87 #define FW_MAX_SIZE             0x400000
88 #define FW_FILE_NAME            "slof.bin"
89 #define FW_OVERHEAD             0x2800000
90 #define KERNEL_LOAD_ADDR        FW_MAX_SIZE
91 
92 #define MIN_RMA_SLOF            128UL
93 
94 #define PHANDLE_XICP            0x00001111
95 
96 #define HTAB_SIZE(spapr)        (1ULL << ((spapr)->htab_shift))
97 
98 static XICSState *try_create_xics(const char *type, int nr_servers,
99                                   int nr_irqs, Error **errp)
100 {
101     Error *err = NULL;
102     DeviceState *dev;
103 
104     dev = qdev_create(NULL, type);
105     qdev_prop_set_uint32(dev, "nr_servers", nr_servers);
106     qdev_prop_set_uint32(dev, "nr_irqs", nr_irqs);
107     object_property_set_bool(OBJECT(dev), true, "realized", &err);
108     if (err) {
109         error_propagate(errp, err);
110         object_unparent(OBJECT(dev));
111         return NULL;
112     }
113     return XICS_COMMON(dev);
114 }
115 
116 static XICSState *xics_system_init(MachineState *machine,
117                                    int nr_servers, int nr_irqs, Error **errp)
118 {
119     XICSState *xics = NULL;
120 
121     if (kvm_enabled()) {
122         Error *err = NULL;
123 
124         if (machine_kernel_irqchip_allowed(machine)) {
125             xics = try_create_xics(TYPE_XICS_SPAPR_KVM, nr_servers, nr_irqs,
126                                    &err);
127         }
128         if (machine_kernel_irqchip_required(machine) && !xics) {
129             error_reportf_err(err,
130                               "kernel_irqchip requested but unavailable: ");
131         } else {
132             error_free(err);
133         }
134     }
135 
136     if (!xics) {
137         xics = try_create_xics(TYPE_XICS_SPAPR, nr_servers, nr_irqs, errp);
138     }
139 
140     return xics;
141 }
142 
143 static int spapr_fixup_cpu_smt_dt(void *fdt, int offset, PowerPCCPU *cpu,
144                                   int smt_threads)
145 {
146     int i, ret = 0;
147     uint32_t servers_prop[smt_threads];
148     uint32_t gservers_prop[smt_threads * 2];
149     int index = ppc_get_vcpu_dt_id(cpu);
150 
151     if (cpu->cpu_version) {
152         ret = fdt_setprop_cell(fdt, offset, "cpu-version", cpu->cpu_version);
153         if (ret < 0) {
154             return ret;
155         }
156     }
157 
158     /* Build interrupt servers and gservers properties */
159     for (i = 0; i < smt_threads; i++) {
160         servers_prop[i] = cpu_to_be32(index + i);
161         /* Hack, direct the group queues back to cpu 0 */
162         gservers_prop[i*2] = cpu_to_be32(index + i);
163         gservers_prop[i*2 + 1] = 0;
164     }
165     ret = fdt_setprop(fdt, offset, "ibm,ppc-interrupt-server#s",
166                       servers_prop, sizeof(servers_prop));
167     if (ret < 0) {
168         return ret;
169     }
170     ret = fdt_setprop(fdt, offset, "ibm,ppc-interrupt-gserver#s",
171                       gservers_prop, sizeof(gservers_prop));
172 
173     return ret;
174 }
175 
176 static int spapr_fixup_cpu_numa_dt(void *fdt, int offset, CPUState *cs)
177 {
178     int ret = 0;
179     PowerPCCPU *cpu = POWERPC_CPU(cs);
180     int index = ppc_get_vcpu_dt_id(cpu);
181     uint32_t associativity[] = {cpu_to_be32(0x5),
182                                 cpu_to_be32(0x0),
183                                 cpu_to_be32(0x0),
184                                 cpu_to_be32(0x0),
185                                 cpu_to_be32(cs->numa_node),
186                                 cpu_to_be32(index)};
187 
188     /* Advertise NUMA via ibm,associativity */
189     if (nb_numa_nodes > 1) {
190         ret = fdt_setprop(fdt, offset, "ibm,associativity", associativity,
191                           sizeof(associativity));
192     }
193 
194     return ret;
195 }
196 
197 static int spapr_fixup_cpu_dt(void *fdt, sPAPRMachineState *spapr)
198 {
199     int ret = 0, offset, cpus_offset;
200     CPUState *cs;
201     char cpu_model[32];
202     int smt = kvmppc_smt_threads();
203     uint32_t pft_size_prop[] = {0, cpu_to_be32(spapr->htab_shift)};
204 
205     CPU_FOREACH(cs) {
206         PowerPCCPU *cpu = POWERPC_CPU(cs);
207         DeviceClass *dc = DEVICE_GET_CLASS(cs);
208         int index = ppc_get_vcpu_dt_id(cpu);
209 
210         if ((index % smt) != 0) {
211             continue;
212         }
213 
214         snprintf(cpu_model, 32, "%s@%x", dc->fw_name, index);
215 
216         cpus_offset = fdt_path_offset(fdt, "/cpus");
217         if (cpus_offset < 0) {
218             cpus_offset = fdt_add_subnode(fdt, fdt_path_offset(fdt, "/"),
219                                           "cpus");
220             if (cpus_offset < 0) {
221                 return cpus_offset;
222             }
223         }
224         offset = fdt_subnode_offset(fdt, cpus_offset, cpu_model);
225         if (offset < 0) {
226             offset = fdt_add_subnode(fdt, cpus_offset, cpu_model);
227             if (offset < 0) {
228                 return offset;
229             }
230         }
231 
232         ret = fdt_setprop(fdt, offset, "ibm,pft-size",
233                           pft_size_prop, sizeof(pft_size_prop));
234         if (ret < 0) {
235             return ret;
236         }
237 
238         ret = spapr_fixup_cpu_numa_dt(fdt, offset, cs);
239         if (ret < 0) {
240             return ret;
241         }
242 
243         ret = spapr_fixup_cpu_smt_dt(fdt, offset, cpu,
244                                      ppc_get_compat_smt_threads(cpu));
245         if (ret < 0) {
246             return ret;
247         }
248     }
249     return ret;
250 }
251 
252 
253 static size_t create_page_sizes_prop(CPUPPCState *env, uint32_t *prop,
254                                      size_t maxsize)
255 {
256     size_t maxcells = maxsize / sizeof(uint32_t);
257     int i, j, count;
258     uint32_t *p = prop;
259 
260     for (i = 0; i < PPC_PAGE_SIZES_MAX_SZ; i++) {
261         struct ppc_one_seg_page_size *sps = &env->sps.sps[i];
262 
263         if (!sps->page_shift) {
264             break;
265         }
266         for (count = 0; count < PPC_PAGE_SIZES_MAX_SZ; count++) {
267             if (sps->enc[count].page_shift == 0) {
268                 break;
269             }
270         }
271         if ((p - prop) >= (maxcells - 3 - count * 2)) {
272             break;
273         }
274         *(p++) = cpu_to_be32(sps->page_shift);
275         *(p++) = cpu_to_be32(sps->slb_enc);
276         *(p++) = cpu_to_be32(count);
277         for (j = 0; j < count; j++) {
278             *(p++) = cpu_to_be32(sps->enc[j].page_shift);
279             *(p++) = cpu_to_be32(sps->enc[j].pte_enc);
280         }
281     }
282 
283     return (p - prop) * sizeof(uint32_t);
284 }
285 
286 static hwaddr spapr_node0_size(void)
287 {
288     MachineState *machine = MACHINE(qdev_get_machine());
289 
290     if (nb_numa_nodes) {
291         int i;
292         for (i = 0; i < nb_numa_nodes; ++i) {
293             if (numa_info[i].node_mem) {
294                 return MIN(pow2floor(numa_info[i].node_mem),
295                            machine->ram_size);
296             }
297         }
298     }
299     return machine->ram_size;
300 }
301 
302 #define _FDT(exp) \
303     do { \
304         int ret = (exp);                                           \
305         if (ret < 0) {                                             \
306             fprintf(stderr, "qemu: error creating device tree: %s: %s\n", \
307                     #exp, fdt_strerror(ret));                      \
308             exit(1);                                               \
309         }                                                          \
310     } while (0)
311 
312 static void add_str(GString *s, const gchar *s1)
313 {
314     g_string_append_len(s, s1, strlen(s1) + 1);
315 }
316 
317 static void *spapr_create_fdt_skel(hwaddr initrd_base,
318                                    hwaddr initrd_size,
319                                    hwaddr kernel_size,
320                                    bool little_endian,
321                                    const char *kernel_cmdline,
322                                    uint32_t epow_irq)
323 {
324     void *fdt;
325     uint32_t start_prop = cpu_to_be32(initrd_base);
326     uint32_t end_prop = cpu_to_be32(initrd_base + initrd_size);
327     GString *hypertas = g_string_sized_new(256);
328     GString *qemu_hypertas = g_string_sized_new(256);
329     uint32_t refpoints[] = {cpu_to_be32(0x4), cpu_to_be32(0x4)};
330     uint32_t interrupt_server_ranges_prop[] = {0, cpu_to_be32(max_cpus)};
331     unsigned char vec5[] = {0x0, 0x0, 0x0, 0x0, 0x0, 0x80};
332     char *buf;
333 
334     add_str(hypertas, "hcall-pft");
335     add_str(hypertas, "hcall-term");
336     add_str(hypertas, "hcall-dabr");
337     add_str(hypertas, "hcall-interrupt");
338     add_str(hypertas, "hcall-tce");
339     add_str(hypertas, "hcall-vio");
340     add_str(hypertas, "hcall-splpar");
341     add_str(hypertas, "hcall-bulk");
342     add_str(hypertas, "hcall-set-mode");
343     add_str(hypertas, "hcall-sprg0");
344     add_str(hypertas, "hcall-copy");
345     add_str(hypertas, "hcall-debug");
346     add_str(qemu_hypertas, "hcall-memop1");
347 
348     fdt = g_malloc0(FDT_MAX_SIZE);
349     _FDT((fdt_create(fdt, FDT_MAX_SIZE)));
350 
351     if (kernel_size) {
352         _FDT((fdt_add_reservemap_entry(fdt, KERNEL_LOAD_ADDR, kernel_size)));
353     }
354     if (initrd_size) {
355         _FDT((fdt_add_reservemap_entry(fdt, initrd_base, initrd_size)));
356     }
357     _FDT((fdt_finish_reservemap(fdt)));
358 
359     /* Root node */
360     _FDT((fdt_begin_node(fdt, "")));
361     _FDT((fdt_property_string(fdt, "device_type", "chrp")));
362     _FDT((fdt_property_string(fdt, "model", "IBM pSeries (emulated by qemu)")));
363     _FDT((fdt_property_string(fdt, "compatible", "qemu,pseries")));
364 
365     /*
366      * Add info to guest to indentify which host is it being run on
367      * and what is the uuid of the guest
368      */
369     if (kvmppc_get_host_model(&buf)) {
370         _FDT((fdt_property_string(fdt, "host-model", buf)));
371         g_free(buf);
372     }
373     if (kvmppc_get_host_serial(&buf)) {
374         _FDT((fdt_property_string(fdt, "host-serial", buf)));
375         g_free(buf);
376     }
377 
378     buf = g_strdup_printf(UUID_FMT, qemu_uuid[0], qemu_uuid[1],
379                           qemu_uuid[2], qemu_uuid[3], qemu_uuid[4],
380                           qemu_uuid[5], qemu_uuid[6], qemu_uuid[7],
381                           qemu_uuid[8], qemu_uuid[9], qemu_uuid[10],
382                           qemu_uuid[11], qemu_uuid[12], qemu_uuid[13],
383                           qemu_uuid[14], qemu_uuid[15]);
384 
385     _FDT((fdt_property_string(fdt, "vm,uuid", buf)));
386     if (qemu_uuid_set) {
387         _FDT((fdt_property_string(fdt, "system-id", buf)));
388     }
389     g_free(buf);
390 
391     if (qemu_get_vm_name()) {
392         _FDT((fdt_property_string(fdt, "ibm,partition-name",
393                                   qemu_get_vm_name())));
394     }
395 
396     _FDT((fdt_property_cell(fdt, "#address-cells", 0x2)));
397     _FDT((fdt_property_cell(fdt, "#size-cells", 0x2)));
398 
399     /* /chosen */
400     _FDT((fdt_begin_node(fdt, "chosen")));
401 
402     /* Set Form1_affinity */
403     _FDT((fdt_property(fdt, "ibm,architecture-vec-5", vec5, sizeof(vec5))));
404 
405     _FDT((fdt_property_string(fdt, "bootargs", kernel_cmdline)));
406     _FDT((fdt_property(fdt, "linux,initrd-start",
407                        &start_prop, sizeof(start_prop))));
408     _FDT((fdt_property(fdt, "linux,initrd-end",
409                        &end_prop, sizeof(end_prop))));
410     if (kernel_size) {
411         uint64_t kprop[2] = { cpu_to_be64(KERNEL_LOAD_ADDR),
412                               cpu_to_be64(kernel_size) };
413 
414         _FDT((fdt_property(fdt, "qemu,boot-kernel", &kprop, sizeof(kprop))));
415         if (little_endian) {
416             _FDT((fdt_property(fdt, "qemu,boot-kernel-le", NULL, 0)));
417         }
418     }
419     if (boot_menu) {
420         _FDT((fdt_property_cell(fdt, "qemu,boot-menu", boot_menu)));
421     }
422     _FDT((fdt_property_cell(fdt, "qemu,graphic-width", graphic_width)));
423     _FDT((fdt_property_cell(fdt, "qemu,graphic-height", graphic_height)));
424     _FDT((fdt_property_cell(fdt, "qemu,graphic-depth", graphic_depth)));
425 
426     _FDT((fdt_end_node(fdt)));
427 
428     /* RTAS */
429     _FDT((fdt_begin_node(fdt, "rtas")));
430 
431     if (!kvm_enabled() || kvmppc_spapr_use_multitce()) {
432         add_str(hypertas, "hcall-multi-tce");
433     }
434     _FDT((fdt_property(fdt, "ibm,hypertas-functions", hypertas->str,
435                        hypertas->len)));
436     g_string_free(hypertas, TRUE);
437     _FDT((fdt_property(fdt, "qemu,hypertas-functions", qemu_hypertas->str,
438                        qemu_hypertas->len)));
439     g_string_free(qemu_hypertas, TRUE);
440 
441     _FDT((fdt_property(fdt, "ibm,associativity-reference-points",
442         refpoints, sizeof(refpoints))));
443 
444     _FDT((fdt_property_cell(fdt, "rtas-error-log-max", RTAS_ERROR_LOG_MAX)));
445     _FDT((fdt_property_cell(fdt, "rtas-event-scan-rate",
446                             RTAS_EVENT_SCAN_RATE)));
447 
448     if (msi_nonbroken) {
449         _FDT((fdt_property(fdt, "ibm,change-msix-capable", NULL, 0)));
450     }
451 
452     /*
453      * According to PAPR, rtas ibm,os-term does not guarantee a return
454      * back to the guest cpu.
455      *
456      * While an additional ibm,extended-os-term property indicates that
457      * rtas call return will always occur. Set this property.
458      */
459     _FDT((fdt_property(fdt, "ibm,extended-os-term", NULL, 0)));
460 
461     _FDT((fdt_end_node(fdt)));
462 
463     /* interrupt controller */
464     _FDT((fdt_begin_node(fdt, "interrupt-controller")));
465 
466     _FDT((fdt_property_string(fdt, "device_type",
467                               "PowerPC-External-Interrupt-Presentation")));
468     _FDT((fdt_property_string(fdt, "compatible", "IBM,ppc-xicp")));
469     _FDT((fdt_property(fdt, "interrupt-controller", NULL, 0)));
470     _FDT((fdt_property(fdt, "ibm,interrupt-server-ranges",
471                        interrupt_server_ranges_prop,
472                        sizeof(interrupt_server_ranges_prop))));
473     _FDT((fdt_property_cell(fdt, "#interrupt-cells", 2)));
474     _FDT((fdt_property_cell(fdt, "linux,phandle", PHANDLE_XICP)));
475     _FDT((fdt_property_cell(fdt, "phandle", PHANDLE_XICP)));
476 
477     _FDT((fdt_end_node(fdt)));
478 
479     /* vdevice */
480     _FDT((fdt_begin_node(fdt, "vdevice")));
481 
482     _FDT((fdt_property_string(fdt, "device_type", "vdevice")));
483     _FDT((fdt_property_string(fdt, "compatible", "IBM,vdevice")));
484     _FDT((fdt_property_cell(fdt, "#address-cells", 0x1)));
485     _FDT((fdt_property_cell(fdt, "#size-cells", 0x0)));
486     _FDT((fdt_property_cell(fdt, "#interrupt-cells", 0x2)));
487     _FDT((fdt_property(fdt, "interrupt-controller", NULL, 0)));
488 
489     _FDT((fdt_end_node(fdt)));
490 
491     /* event-sources */
492     spapr_events_fdt_skel(fdt, epow_irq);
493 
494     /* /hypervisor node */
495     if (kvm_enabled()) {
496         uint8_t hypercall[16];
497 
498         /* indicate KVM hypercall interface */
499         _FDT((fdt_begin_node(fdt, "hypervisor")));
500         _FDT((fdt_property_string(fdt, "compatible", "linux,kvm")));
501         if (kvmppc_has_cap_fixup_hcalls()) {
502             /*
503              * Older KVM versions with older guest kernels were broken with the
504              * magic page, don't allow the guest to map it.
505              */
506             if (!kvmppc_get_hypercall(first_cpu->env_ptr, hypercall,
507                                       sizeof(hypercall))) {
508                 _FDT((fdt_property(fdt, "hcall-instructions", hypercall,
509                                    sizeof(hypercall))));
510             }
511         }
512         _FDT((fdt_end_node(fdt)));
513     }
514 
515     _FDT((fdt_end_node(fdt))); /* close root node */
516     _FDT((fdt_finish(fdt)));
517 
518     return fdt;
519 }
520 
521 static int spapr_populate_memory_node(void *fdt, int nodeid, hwaddr start,
522                                        hwaddr size)
523 {
524     uint32_t associativity[] = {
525         cpu_to_be32(0x4), /* length */
526         cpu_to_be32(0x0), cpu_to_be32(0x0),
527         cpu_to_be32(0x0), cpu_to_be32(nodeid)
528     };
529     char mem_name[32];
530     uint64_t mem_reg_property[2];
531     int off;
532 
533     mem_reg_property[0] = cpu_to_be64(start);
534     mem_reg_property[1] = cpu_to_be64(size);
535 
536     sprintf(mem_name, "memory@" TARGET_FMT_lx, start);
537     off = fdt_add_subnode(fdt, 0, mem_name);
538     _FDT(off);
539     _FDT((fdt_setprop_string(fdt, off, "device_type", "memory")));
540     _FDT((fdt_setprop(fdt, off, "reg", mem_reg_property,
541                       sizeof(mem_reg_property))));
542     _FDT((fdt_setprop(fdt, off, "ibm,associativity", associativity,
543                       sizeof(associativity))));
544     return off;
545 }
546 
547 static int spapr_populate_memory(sPAPRMachineState *spapr, void *fdt)
548 {
549     MachineState *machine = MACHINE(spapr);
550     hwaddr mem_start, node_size;
551     int i, nb_nodes = nb_numa_nodes;
552     NodeInfo *nodes = numa_info;
553     NodeInfo ramnode;
554 
555     /* No NUMA nodes, assume there is just one node with whole RAM */
556     if (!nb_numa_nodes) {
557         nb_nodes = 1;
558         ramnode.node_mem = machine->ram_size;
559         nodes = &ramnode;
560     }
561 
562     for (i = 0, mem_start = 0; i < nb_nodes; ++i) {
563         if (!nodes[i].node_mem) {
564             continue;
565         }
566         if (mem_start >= machine->ram_size) {
567             node_size = 0;
568         } else {
569             node_size = nodes[i].node_mem;
570             if (node_size > machine->ram_size - mem_start) {
571                 node_size = machine->ram_size - mem_start;
572             }
573         }
574         if (!mem_start) {
575             /* ppc_spapr_init() checks for rma_size <= node0_size already */
576             spapr_populate_memory_node(fdt, i, 0, spapr->rma_size);
577             mem_start += spapr->rma_size;
578             node_size -= spapr->rma_size;
579         }
580         for ( ; node_size; ) {
581             hwaddr sizetmp = pow2floor(node_size);
582 
583             /* mem_start != 0 here */
584             if (ctzl(mem_start) < ctzl(sizetmp)) {
585                 sizetmp = 1ULL << ctzl(mem_start);
586             }
587 
588             spapr_populate_memory_node(fdt, i, mem_start, sizetmp);
589             node_size -= sizetmp;
590             mem_start += sizetmp;
591         }
592     }
593 
594     return 0;
595 }
596 
597 static void spapr_populate_cpu_dt(CPUState *cs, void *fdt, int offset,
598                                   sPAPRMachineState *spapr)
599 {
600     PowerPCCPU *cpu = POWERPC_CPU(cs);
601     CPUPPCState *env = &cpu->env;
602     PowerPCCPUClass *pcc = POWERPC_CPU_GET_CLASS(cs);
603     int index = ppc_get_vcpu_dt_id(cpu);
604     uint32_t segs[] = {cpu_to_be32(28), cpu_to_be32(40),
605                        0xffffffff, 0xffffffff};
606     uint32_t tbfreq = kvm_enabled() ? kvmppc_get_tbfreq()
607         : SPAPR_TIMEBASE_FREQ;
608     uint32_t cpufreq = kvm_enabled() ? kvmppc_get_clockfreq() : 1000000000;
609     uint32_t page_sizes_prop[64];
610     size_t page_sizes_prop_size;
611     uint32_t vcpus_per_socket = smp_threads * smp_cores;
612     uint32_t pft_size_prop[] = {0, cpu_to_be32(spapr->htab_shift)};
613     sPAPRDRConnector *drc;
614     sPAPRDRConnectorClass *drck;
615     int drc_index;
616 
617     drc = spapr_dr_connector_by_id(SPAPR_DR_CONNECTOR_TYPE_CPU, index);
618     if (drc) {
619         drck = SPAPR_DR_CONNECTOR_GET_CLASS(drc);
620         drc_index = drck->get_index(drc);
621         _FDT((fdt_setprop_cell(fdt, offset, "ibm,my-drc-index", drc_index)));
622     }
623 
624     /* Note: we keep CI large pages off for now because a 64K capable guest
625      * provisioned with large pages might otherwise try to map a qemu
626      * framebuffer (or other kind of memory mapped PCI BAR) using 64K pages
627      * even if that qemu runs on a 4k host.
628      *
629      * We can later add this bit back when we are confident this is not
630      * an issue (!HV KVM or 64K host)
631      */
632     uint8_t pa_features_206[] = { 6, 0,
633         0xf6, 0x1f, 0xc7, 0x00, 0x80, 0xc0 };
634     uint8_t pa_features_207[] = { 24, 0,
635         0xf6, 0x1f, 0xc7, 0xc0, 0x80, 0xf0,
636         0x80, 0x00, 0x00, 0x00, 0x00, 0x00,
637         0x00, 0x00, 0x00, 0x00, 0x80, 0x00,
638         0x80, 0x00, 0x80, 0x00, 0x80, 0x00 };
639     uint8_t *pa_features;
640     size_t pa_size;
641 
642     _FDT((fdt_setprop_cell(fdt, offset, "reg", index)));
643     _FDT((fdt_setprop_string(fdt, offset, "device_type", "cpu")));
644 
645     _FDT((fdt_setprop_cell(fdt, offset, "cpu-version", env->spr[SPR_PVR])));
646     _FDT((fdt_setprop_cell(fdt, offset, "d-cache-block-size",
647                            env->dcache_line_size)));
648     _FDT((fdt_setprop_cell(fdt, offset, "d-cache-line-size",
649                            env->dcache_line_size)));
650     _FDT((fdt_setprop_cell(fdt, offset, "i-cache-block-size",
651                            env->icache_line_size)));
652     _FDT((fdt_setprop_cell(fdt, offset, "i-cache-line-size",
653                            env->icache_line_size)));
654 
655     if (pcc->l1_dcache_size) {
656         _FDT((fdt_setprop_cell(fdt, offset, "d-cache-size",
657                                pcc->l1_dcache_size)));
658     } else {
659         fprintf(stderr, "Warning: Unknown L1 dcache size for cpu\n");
660     }
661     if (pcc->l1_icache_size) {
662         _FDT((fdt_setprop_cell(fdt, offset, "i-cache-size",
663                                pcc->l1_icache_size)));
664     } else {
665         fprintf(stderr, "Warning: Unknown L1 icache size for cpu\n");
666     }
667 
668     _FDT((fdt_setprop_cell(fdt, offset, "timebase-frequency", tbfreq)));
669     _FDT((fdt_setprop_cell(fdt, offset, "clock-frequency", cpufreq)));
670     _FDT((fdt_setprop_cell(fdt, offset, "slb-size", env->slb_nr)));
671     _FDT((fdt_setprop_cell(fdt, offset, "ibm,slb-size", env->slb_nr)));
672     _FDT((fdt_setprop_string(fdt, offset, "status", "okay")));
673     _FDT((fdt_setprop(fdt, offset, "64-bit", NULL, 0)));
674 
675     if (env->spr_cb[SPR_PURR].oea_read) {
676         _FDT((fdt_setprop(fdt, offset, "ibm,purr", NULL, 0)));
677     }
678 
679     if (env->mmu_model & POWERPC_MMU_1TSEG) {
680         _FDT((fdt_setprop(fdt, offset, "ibm,processor-segment-sizes",
681                           segs, sizeof(segs))));
682     }
683 
684     /* Advertise VMX/VSX (vector extensions) if available
685      *   0 / no property == no vector extensions
686      *   1               == VMX / Altivec available
687      *   2               == VSX available */
688     if (env->insns_flags & PPC_ALTIVEC) {
689         uint32_t vmx = (env->insns_flags2 & PPC2_VSX) ? 2 : 1;
690 
691         _FDT((fdt_setprop_cell(fdt, offset, "ibm,vmx", vmx)));
692     }
693 
694     /* Advertise DFP (Decimal Floating Point) if available
695      *   0 / no property == no DFP
696      *   1               == DFP available */
697     if (env->insns_flags2 & PPC2_DFP) {
698         _FDT((fdt_setprop_cell(fdt, offset, "ibm,dfp", 1)));
699     }
700 
701     page_sizes_prop_size = create_page_sizes_prop(env, page_sizes_prop,
702                                                   sizeof(page_sizes_prop));
703     if (page_sizes_prop_size) {
704         _FDT((fdt_setprop(fdt, offset, "ibm,segment-page-sizes",
705                           page_sizes_prop, page_sizes_prop_size)));
706     }
707 
708     /* Do the ibm,pa-features property, adjust it for ci-large-pages */
709     if (env->mmu_model == POWERPC_MMU_2_06) {
710         pa_features = pa_features_206;
711         pa_size = sizeof(pa_features_206);
712     } else /* env->mmu_model == POWERPC_MMU_2_07 */ {
713         pa_features = pa_features_207;
714         pa_size = sizeof(pa_features_207);
715     }
716     if (env->ci_large_pages) {
717         pa_features[3] |= 0x20;
718     }
719     _FDT((fdt_setprop(fdt, offset, "ibm,pa-features", pa_features, pa_size)));
720 
721     _FDT((fdt_setprop_cell(fdt, offset, "ibm,chip-id",
722                            cs->cpu_index / vcpus_per_socket)));
723 
724     _FDT((fdt_setprop(fdt, offset, "ibm,pft-size",
725                       pft_size_prop, sizeof(pft_size_prop))));
726 
727     _FDT(spapr_fixup_cpu_numa_dt(fdt, offset, cs));
728 
729     _FDT(spapr_fixup_cpu_smt_dt(fdt, offset, cpu,
730                                 ppc_get_compat_smt_threads(cpu)));
731 }
732 
733 static void spapr_populate_cpus_dt_node(void *fdt, sPAPRMachineState *spapr)
734 {
735     CPUState *cs;
736     int cpus_offset;
737     char *nodename;
738     int smt = kvmppc_smt_threads();
739 
740     cpus_offset = fdt_add_subnode(fdt, 0, "cpus");
741     _FDT(cpus_offset);
742     _FDT((fdt_setprop_cell(fdt, cpus_offset, "#address-cells", 0x1)));
743     _FDT((fdt_setprop_cell(fdt, cpus_offset, "#size-cells", 0x0)));
744 
745     /*
746      * We walk the CPUs in reverse order to ensure that CPU DT nodes
747      * created by fdt_add_subnode() end up in the right order in FDT
748      * for the guest kernel the enumerate the CPUs correctly.
749      */
750     CPU_FOREACH_REVERSE(cs) {
751         PowerPCCPU *cpu = POWERPC_CPU(cs);
752         int index = ppc_get_vcpu_dt_id(cpu);
753         DeviceClass *dc = DEVICE_GET_CLASS(cs);
754         int offset;
755 
756         if ((index % smt) != 0) {
757             continue;
758         }
759 
760         nodename = g_strdup_printf("%s@%x", dc->fw_name, index);
761         offset = fdt_add_subnode(fdt, cpus_offset, nodename);
762         g_free(nodename);
763         _FDT(offset);
764         spapr_populate_cpu_dt(cs, fdt, offset, spapr);
765     }
766 
767 }
768 
769 /*
770  * Adds ibm,dynamic-reconfiguration-memory node.
771  * Refer to docs/specs/ppc-spapr-hotplug.txt for the documentation
772  * of this device tree node.
773  */
774 static int spapr_populate_drconf_memory(sPAPRMachineState *spapr, void *fdt)
775 {
776     MachineState *machine = MACHINE(spapr);
777     int ret, i, offset;
778     uint64_t lmb_size = SPAPR_MEMORY_BLOCK_SIZE;
779     uint32_t prop_lmb_size[] = {0, cpu_to_be32(lmb_size)};
780     uint32_t hotplug_lmb_start = spapr->hotplug_memory.base / lmb_size;
781     uint32_t nr_lmbs = (spapr->hotplug_memory.base +
782                        memory_region_size(&spapr->hotplug_memory.mr)) /
783                        lmb_size;
784     uint32_t *int_buf, *cur_index, buf_len;
785     int nr_nodes = nb_numa_nodes ? nb_numa_nodes : 1;
786 
787     /*
788      * Don't create the node if there is no hotpluggable memory
789      */
790     if (machine->ram_size == machine->maxram_size) {
791         return 0;
792     }
793 
794     /*
795      * Allocate enough buffer size to fit in ibm,dynamic-memory
796      * or ibm,associativity-lookup-arrays
797      */
798     buf_len = MAX(nr_lmbs * SPAPR_DR_LMB_LIST_ENTRY_SIZE + 1, nr_nodes * 4 + 2)
799               * sizeof(uint32_t);
800     cur_index = int_buf = g_malloc0(buf_len);
801 
802     offset = fdt_add_subnode(fdt, 0, "ibm,dynamic-reconfiguration-memory");
803 
804     ret = fdt_setprop(fdt, offset, "ibm,lmb-size", prop_lmb_size,
805                     sizeof(prop_lmb_size));
806     if (ret < 0) {
807         goto out;
808     }
809 
810     ret = fdt_setprop_cell(fdt, offset, "ibm,memory-flags-mask", 0xff);
811     if (ret < 0) {
812         goto out;
813     }
814 
815     ret = fdt_setprop_cell(fdt, offset, "ibm,memory-preservation-time", 0x0);
816     if (ret < 0) {
817         goto out;
818     }
819 
820     /* ibm,dynamic-memory */
821     int_buf[0] = cpu_to_be32(nr_lmbs);
822     cur_index++;
823     for (i = 0; i < nr_lmbs; i++) {
824         uint64_t addr = i * lmb_size;
825         uint32_t *dynamic_memory = cur_index;
826 
827         if (i >= hotplug_lmb_start) {
828             sPAPRDRConnector *drc;
829             sPAPRDRConnectorClass *drck;
830 
831             drc = spapr_dr_connector_by_id(SPAPR_DR_CONNECTOR_TYPE_LMB, i);
832             g_assert(drc);
833             drck = SPAPR_DR_CONNECTOR_GET_CLASS(drc);
834 
835             dynamic_memory[0] = cpu_to_be32(addr >> 32);
836             dynamic_memory[1] = cpu_to_be32(addr & 0xffffffff);
837             dynamic_memory[2] = cpu_to_be32(drck->get_index(drc));
838             dynamic_memory[3] = cpu_to_be32(0); /* reserved */
839             dynamic_memory[4] = cpu_to_be32(numa_get_node(addr, NULL));
840             if (memory_region_present(get_system_memory(), addr)) {
841                 dynamic_memory[5] = cpu_to_be32(SPAPR_LMB_FLAGS_ASSIGNED);
842             } else {
843                 dynamic_memory[5] = cpu_to_be32(0);
844             }
845         } else {
846             /*
847              * LMB information for RMA, boot time RAM and gap b/n RAM and
848              * hotplug memory region -- all these are marked as reserved
849              * and as having no valid DRC.
850              */
851             dynamic_memory[0] = cpu_to_be32(addr >> 32);
852             dynamic_memory[1] = cpu_to_be32(addr & 0xffffffff);
853             dynamic_memory[2] = cpu_to_be32(0);
854             dynamic_memory[3] = cpu_to_be32(0); /* reserved */
855             dynamic_memory[4] = cpu_to_be32(-1);
856             dynamic_memory[5] = cpu_to_be32(SPAPR_LMB_FLAGS_RESERVED |
857                                             SPAPR_LMB_FLAGS_DRC_INVALID);
858         }
859 
860         cur_index += SPAPR_DR_LMB_LIST_ENTRY_SIZE;
861     }
862     ret = fdt_setprop(fdt, offset, "ibm,dynamic-memory", int_buf, buf_len);
863     if (ret < 0) {
864         goto out;
865     }
866 
867     /* ibm,associativity-lookup-arrays */
868     cur_index = int_buf;
869     int_buf[0] = cpu_to_be32(nr_nodes);
870     int_buf[1] = cpu_to_be32(4); /* Number of entries per associativity list */
871     cur_index += 2;
872     for (i = 0; i < nr_nodes; i++) {
873         uint32_t associativity[] = {
874             cpu_to_be32(0x0),
875             cpu_to_be32(0x0),
876             cpu_to_be32(0x0),
877             cpu_to_be32(i)
878         };
879         memcpy(cur_index, associativity, sizeof(associativity));
880         cur_index += 4;
881     }
882     ret = fdt_setprop(fdt, offset, "ibm,associativity-lookup-arrays", int_buf,
883             (cur_index - int_buf) * sizeof(uint32_t));
884 out:
885     g_free(int_buf);
886     return ret;
887 }
888 
889 int spapr_h_cas_compose_response(sPAPRMachineState *spapr,
890                                  target_ulong addr, target_ulong size,
891                                  bool cpu_update, bool memory_update)
892 {
893     void *fdt, *fdt_skel;
894     sPAPRDeviceTreeUpdateHeader hdr = { .version_id = 1 };
895     sPAPRMachineClass *smc = SPAPR_MACHINE_GET_CLASS(qdev_get_machine());
896 
897     size -= sizeof(hdr);
898 
899     /* Create sceleton */
900     fdt_skel = g_malloc0(size);
901     _FDT((fdt_create(fdt_skel, size)));
902     _FDT((fdt_begin_node(fdt_skel, "")));
903     _FDT((fdt_end_node(fdt_skel)));
904     _FDT((fdt_finish(fdt_skel)));
905     fdt = g_malloc0(size);
906     _FDT((fdt_open_into(fdt_skel, fdt, size)));
907     g_free(fdt_skel);
908 
909     /* Fixup cpu nodes */
910     if (cpu_update) {
911         _FDT((spapr_fixup_cpu_dt(fdt, spapr)));
912     }
913 
914     /* Generate ibm,dynamic-reconfiguration-memory node if required */
915     if (memory_update && smc->dr_lmb_enabled) {
916         _FDT((spapr_populate_drconf_memory(spapr, fdt)));
917     }
918 
919     /* Pack resulting tree */
920     _FDT((fdt_pack(fdt)));
921 
922     if (fdt_totalsize(fdt) + sizeof(hdr) > size) {
923         trace_spapr_cas_failed(size);
924         return -1;
925     }
926 
927     cpu_physical_memory_write(addr, &hdr, sizeof(hdr));
928     cpu_physical_memory_write(addr + sizeof(hdr), fdt, fdt_totalsize(fdt));
929     trace_spapr_cas_continue(fdt_totalsize(fdt) + sizeof(hdr));
930     g_free(fdt);
931 
932     return 0;
933 }
934 
935 static void spapr_finalize_fdt(sPAPRMachineState *spapr,
936                                hwaddr fdt_addr,
937                                hwaddr rtas_addr,
938                                hwaddr rtas_size)
939 {
940     MachineState *machine = MACHINE(qdev_get_machine());
941     MachineClass *mc = MACHINE_GET_CLASS(machine);
942     sPAPRMachineClass *smc = SPAPR_MACHINE_GET_CLASS(machine);
943     const char *boot_device = machine->boot_order;
944     int ret, i;
945     size_t cb = 0;
946     char *bootlist;
947     void *fdt;
948     sPAPRPHBState *phb;
949 
950     fdt = g_malloc(FDT_MAX_SIZE);
951 
952     /* open out the base tree into a temp buffer for the final tweaks */
953     _FDT((fdt_open_into(spapr->fdt_skel, fdt, FDT_MAX_SIZE)));
954 
955     ret = spapr_populate_memory(spapr, fdt);
956     if (ret < 0) {
957         fprintf(stderr, "couldn't setup memory nodes in fdt\n");
958         exit(1);
959     }
960 
961     ret = spapr_populate_vdevice(spapr->vio_bus, fdt);
962     if (ret < 0) {
963         fprintf(stderr, "couldn't setup vio devices in fdt\n");
964         exit(1);
965     }
966 
967     if (object_resolve_path_type("", TYPE_SPAPR_RNG, NULL)) {
968         ret = spapr_rng_populate_dt(fdt);
969         if (ret < 0) {
970             fprintf(stderr, "could not set up rng device in the fdt\n");
971             exit(1);
972         }
973     }
974 
975     QLIST_FOREACH(phb, &spapr->phbs, list) {
976         ret = spapr_populate_pci_dt(phb, PHANDLE_XICP, fdt);
977         if (ret < 0) {
978             error_report("couldn't setup PCI devices in fdt");
979             exit(1);
980         }
981     }
982 
983     /* RTAS */
984     ret = spapr_rtas_device_tree_setup(fdt, rtas_addr, rtas_size);
985     if (ret < 0) {
986         fprintf(stderr, "Couldn't set up RTAS device tree properties\n");
987     }
988 
989     /* cpus */
990     spapr_populate_cpus_dt_node(fdt, spapr);
991 
992     bootlist = get_boot_devices_list(&cb, true);
993     if (cb && bootlist) {
994         int offset = fdt_path_offset(fdt, "/chosen");
995         if (offset < 0) {
996             exit(1);
997         }
998         for (i = 0; i < cb; i++) {
999             if (bootlist[i] == '\n') {
1000                 bootlist[i] = ' ';
1001             }
1002 
1003         }
1004         ret = fdt_setprop_string(fdt, offset, "qemu,boot-list", bootlist);
1005     }
1006 
1007     if (boot_device && strlen(boot_device)) {
1008         int offset = fdt_path_offset(fdt, "/chosen");
1009 
1010         if (offset < 0) {
1011             exit(1);
1012         }
1013         fdt_setprop_string(fdt, offset, "qemu,boot-device", boot_device);
1014     }
1015 
1016     if (!spapr->has_graphics) {
1017         spapr_populate_chosen_stdout(fdt, spapr->vio_bus);
1018     }
1019 
1020     if (smc->dr_lmb_enabled) {
1021         _FDT(spapr_drc_populate_dt(fdt, 0, NULL, SPAPR_DR_CONNECTOR_TYPE_LMB));
1022     }
1023 
1024     if (mc->query_hotpluggable_cpus) {
1025         int offset = fdt_path_offset(fdt, "/cpus");
1026         ret = spapr_drc_populate_dt(fdt, offset, NULL,
1027                                     SPAPR_DR_CONNECTOR_TYPE_CPU);
1028         if (ret < 0) {
1029             error_report("Couldn't set up CPU DR device tree properties");
1030             exit(1);
1031         }
1032     }
1033 
1034     _FDT((fdt_pack(fdt)));
1035 
1036     if (fdt_totalsize(fdt) > FDT_MAX_SIZE) {
1037         error_report("FDT too big ! 0x%x bytes (max is 0x%x)",
1038                      fdt_totalsize(fdt), FDT_MAX_SIZE);
1039         exit(1);
1040     }
1041 
1042     qemu_fdt_dumpdtb(fdt, fdt_totalsize(fdt));
1043     cpu_physical_memory_write(fdt_addr, fdt, fdt_totalsize(fdt));
1044 
1045     g_free(bootlist);
1046     g_free(fdt);
1047 }
1048 
1049 static uint64_t translate_kernel_address(void *opaque, uint64_t addr)
1050 {
1051     return (addr & 0x0fffffff) + KERNEL_LOAD_ADDR;
1052 }
1053 
1054 static void emulate_spapr_hypercall(PowerPCCPU *cpu)
1055 {
1056     CPUPPCState *env = &cpu->env;
1057 
1058     if (msr_pr) {
1059         hcall_dprintf("Hypercall made with MSR[PR]=1\n");
1060         env->gpr[3] = H_PRIVILEGE;
1061     } else {
1062         env->gpr[3] = spapr_hypercall(cpu, env->gpr[3], &env->gpr[4]);
1063     }
1064 }
1065 
1066 #define HPTE(_table, _i)   (void *)(((uint64_t *)(_table)) + ((_i) * 2))
1067 #define HPTE_VALID(_hpte)  (tswap64(*((uint64_t *)(_hpte))) & HPTE64_V_VALID)
1068 #define HPTE_DIRTY(_hpte)  (tswap64(*((uint64_t *)(_hpte))) & HPTE64_V_HPTE_DIRTY)
1069 #define CLEAN_HPTE(_hpte)  ((*(uint64_t *)(_hpte)) &= tswap64(~HPTE64_V_HPTE_DIRTY))
1070 #define DIRTY_HPTE(_hpte)  ((*(uint64_t *)(_hpte)) |= tswap64(HPTE64_V_HPTE_DIRTY))
1071 
1072 /*
1073  * Get the fd to access the kernel htab, re-opening it if necessary
1074  */
1075 static int get_htab_fd(sPAPRMachineState *spapr)
1076 {
1077     if (spapr->htab_fd >= 0) {
1078         return spapr->htab_fd;
1079     }
1080 
1081     spapr->htab_fd = kvmppc_get_htab_fd(false);
1082     if (spapr->htab_fd < 0) {
1083         error_report("Unable to open fd for reading hash table from KVM: %s",
1084                      strerror(errno));
1085     }
1086 
1087     return spapr->htab_fd;
1088 }
1089 
1090 static void close_htab_fd(sPAPRMachineState *spapr)
1091 {
1092     if (spapr->htab_fd >= 0) {
1093         close(spapr->htab_fd);
1094     }
1095     spapr->htab_fd = -1;
1096 }
1097 
1098 static int spapr_hpt_shift_for_ramsize(uint64_t ramsize)
1099 {
1100     int shift;
1101 
1102     /* We aim for a hash table of size 1/128 the size of RAM (rounded
1103      * up).  The PAPR recommendation is actually 1/64 of RAM size, but
1104      * that's much more than is needed for Linux guests */
1105     shift = ctz64(pow2ceil(ramsize)) - 7;
1106     shift = MAX(shift, 18); /* Minimum architected size */
1107     shift = MIN(shift, 46); /* Maximum architected size */
1108     return shift;
1109 }
1110 
1111 static void spapr_reallocate_hpt(sPAPRMachineState *spapr, int shift,
1112                                  Error **errp)
1113 {
1114     long rc;
1115 
1116     /* Clean up any HPT info from a previous boot */
1117     g_free(spapr->htab);
1118     spapr->htab = NULL;
1119     spapr->htab_shift = 0;
1120     close_htab_fd(spapr);
1121 
1122     rc = kvmppc_reset_htab(shift);
1123     if (rc < 0) {
1124         /* kernel-side HPT needed, but couldn't allocate one */
1125         error_setg_errno(errp, errno,
1126                          "Failed to allocate KVM HPT of order %d (try smaller maxmem?)",
1127                          shift);
1128         /* This is almost certainly fatal, but if the caller really
1129          * wants to carry on with shift == 0, it's welcome to try */
1130     } else if (rc > 0) {
1131         /* kernel-side HPT allocated */
1132         if (rc != shift) {
1133             error_setg(errp,
1134                        "Requested order %d HPT, but kernel allocated order %ld (try smaller maxmem?)",
1135                        shift, rc);
1136         }
1137 
1138         spapr->htab_shift = shift;
1139         spapr->htab = NULL;
1140     } else {
1141         /* kernel-side HPT not needed, allocate in userspace instead */
1142         size_t size = 1ULL << shift;
1143         int i;
1144 
1145         spapr->htab = qemu_memalign(size, size);
1146         if (!spapr->htab) {
1147             error_setg_errno(errp, errno,
1148                              "Could not allocate HPT of order %d", shift);
1149             return;
1150         }
1151 
1152         memset(spapr->htab, 0, size);
1153         spapr->htab_shift = shift;
1154 
1155         for (i = 0; i < size / HASH_PTE_SIZE_64; i++) {
1156             DIRTY_HPTE(HPTE(spapr->htab, i));
1157         }
1158     }
1159 }
1160 
1161 static int find_unknown_sysbus_device(SysBusDevice *sbdev, void *opaque)
1162 {
1163     bool matched = false;
1164 
1165     if (object_dynamic_cast(OBJECT(sbdev), TYPE_SPAPR_PCI_HOST_BRIDGE)) {
1166         matched = true;
1167     }
1168 
1169     if (!matched) {
1170         error_report("Device %s is not supported by this machine yet.",
1171                      qdev_fw_name(DEVICE(sbdev)));
1172         exit(1);
1173     }
1174 
1175     return 0;
1176 }
1177 
1178 static void ppc_spapr_reset(void)
1179 {
1180     MachineState *machine = MACHINE(qdev_get_machine());
1181     sPAPRMachineState *spapr = SPAPR_MACHINE(machine);
1182     PowerPCCPU *first_ppc_cpu;
1183     uint32_t rtas_limit;
1184 
1185     /* Check for unknown sysbus devices */
1186     foreach_dynamic_sysbus_device(find_unknown_sysbus_device, NULL);
1187 
1188     /* Allocate and/or reset the hash page table */
1189     spapr_reallocate_hpt(spapr,
1190                          spapr_hpt_shift_for_ramsize(machine->maxram_size),
1191                          &error_fatal);
1192 
1193     /* Update the RMA size if necessary */
1194     if (spapr->vrma_adjust) {
1195         spapr->rma_size = kvmppc_rma_size(spapr_node0_size(),
1196                                           spapr->htab_shift);
1197     }
1198 
1199     qemu_devices_reset();
1200 
1201     /*
1202      * We place the device tree and RTAS just below either the top of the RMA,
1203      * or just below 2GB, whichever is lowere, so that it can be
1204      * processed with 32-bit real mode code if necessary
1205      */
1206     rtas_limit = MIN(spapr->rma_size, RTAS_MAX_ADDR);
1207     spapr->rtas_addr = rtas_limit - RTAS_MAX_SIZE;
1208     spapr->fdt_addr = spapr->rtas_addr - FDT_MAX_SIZE;
1209 
1210     /* Load the fdt */
1211     spapr_finalize_fdt(spapr, spapr->fdt_addr, spapr->rtas_addr,
1212                        spapr->rtas_size);
1213 
1214     /* Copy RTAS over */
1215     cpu_physical_memory_write(spapr->rtas_addr, spapr->rtas_blob,
1216                               spapr->rtas_size);
1217 
1218     /* Set up the entry state */
1219     first_ppc_cpu = POWERPC_CPU(first_cpu);
1220     first_ppc_cpu->env.gpr[3] = spapr->fdt_addr;
1221     first_ppc_cpu->env.gpr[5] = 0;
1222     first_cpu->halted = 0;
1223     first_ppc_cpu->env.nip = SPAPR_ENTRY_POINT;
1224 
1225 }
1226 
1227 static void spapr_create_nvram(sPAPRMachineState *spapr)
1228 {
1229     DeviceState *dev = qdev_create(&spapr->vio_bus->bus, "spapr-nvram");
1230     DriveInfo *dinfo = drive_get(IF_PFLASH, 0, 0);
1231 
1232     if (dinfo) {
1233         qdev_prop_set_drive(dev, "drive", blk_by_legacy_dinfo(dinfo),
1234                             &error_fatal);
1235     }
1236 
1237     qdev_init_nofail(dev);
1238 
1239     spapr->nvram = (struct sPAPRNVRAM *)dev;
1240 }
1241 
1242 static void spapr_rtc_create(sPAPRMachineState *spapr)
1243 {
1244     DeviceState *dev = qdev_create(NULL, TYPE_SPAPR_RTC);
1245 
1246     qdev_init_nofail(dev);
1247     spapr->rtc = dev;
1248 
1249     object_property_add_alias(qdev_get_machine(), "rtc-time",
1250                               OBJECT(spapr->rtc), "date", NULL);
1251 }
1252 
1253 /* Returns whether we want to use VGA or not */
1254 static bool spapr_vga_init(PCIBus *pci_bus, Error **errp)
1255 {
1256     switch (vga_interface_type) {
1257     case VGA_NONE:
1258         return false;
1259     case VGA_DEVICE:
1260         return true;
1261     case VGA_STD:
1262     case VGA_VIRTIO:
1263         return pci_vga_init(pci_bus) != NULL;
1264     default:
1265         error_setg(errp,
1266                    "Unsupported VGA mode, only -vga std or -vga virtio is supported");
1267         return false;
1268     }
1269 }
1270 
1271 static int spapr_post_load(void *opaque, int version_id)
1272 {
1273     sPAPRMachineState *spapr = (sPAPRMachineState *)opaque;
1274     int err = 0;
1275 
1276     /* In earlier versions, there was no separate qdev for the PAPR
1277      * RTC, so the RTC offset was stored directly in sPAPREnvironment.
1278      * So when migrating from those versions, poke the incoming offset
1279      * value into the RTC device */
1280     if (version_id < 3) {
1281         err = spapr_rtc_import_offset(spapr->rtc, spapr->rtc_offset);
1282     }
1283 
1284     return err;
1285 }
1286 
1287 static bool version_before_3(void *opaque, int version_id)
1288 {
1289     return version_id < 3;
1290 }
1291 
1292 static const VMStateDescription vmstate_spapr = {
1293     .name = "spapr",
1294     .version_id = 3,
1295     .minimum_version_id = 1,
1296     .post_load = spapr_post_load,
1297     .fields = (VMStateField[]) {
1298         /* used to be @next_irq */
1299         VMSTATE_UNUSED_BUFFER(version_before_3, 0, 4),
1300 
1301         /* RTC offset */
1302         VMSTATE_UINT64_TEST(rtc_offset, sPAPRMachineState, version_before_3),
1303 
1304         VMSTATE_PPC_TIMEBASE_V(tb, sPAPRMachineState, 2),
1305         VMSTATE_END_OF_LIST()
1306     },
1307 };
1308 
1309 static int htab_save_setup(QEMUFile *f, void *opaque)
1310 {
1311     sPAPRMachineState *spapr = opaque;
1312 
1313     /* "Iteration" header */
1314     qemu_put_be32(f, spapr->htab_shift);
1315 
1316     if (spapr->htab) {
1317         spapr->htab_save_index = 0;
1318         spapr->htab_first_pass = true;
1319     } else {
1320         assert(kvm_enabled());
1321     }
1322 
1323 
1324     return 0;
1325 }
1326 
1327 static void htab_save_first_pass(QEMUFile *f, sPAPRMachineState *spapr,
1328                                  int64_t max_ns)
1329 {
1330     bool has_timeout = max_ns != -1;
1331     int htabslots = HTAB_SIZE(spapr) / HASH_PTE_SIZE_64;
1332     int index = spapr->htab_save_index;
1333     int64_t starttime = qemu_clock_get_ns(QEMU_CLOCK_REALTIME);
1334 
1335     assert(spapr->htab_first_pass);
1336 
1337     do {
1338         int chunkstart;
1339 
1340         /* Consume invalid HPTEs */
1341         while ((index < htabslots)
1342                && !HPTE_VALID(HPTE(spapr->htab, index))) {
1343             index++;
1344             CLEAN_HPTE(HPTE(spapr->htab, index));
1345         }
1346 
1347         /* Consume valid HPTEs */
1348         chunkstart = index;
1349         while ((index < htabslots) && (index - chunkstart < USHRT_MAX)
1350                && HPTE_VALID(HPTE(spapr->htab, index))) {
1351             index++;
1352             CLEAN_HPTE(HPTE(spapr->htab, index));
1353         }
1354 
1355         if (index > chunkstart) {
1356             int n_valid = index - chunkstart;
1357 
1358             qemu_put_be32(f, chunkstart);
1359             qemu_put_be16(f, n_valid);
1360             qemu_put_be16(f, 0);
1361             qemu_put_buffer(f, HPTE(spapr->htab, chunkstart),
1362                             HASH_PTE_SIZE_64 * n_valid);
1363 
1364             if (has_timeout &&
1365                 (qemu_clock_get_ns(QEMU_CLOCK_REALTIME) - starttime) > max_ns) {
1366                 break;
1367             }
1368         }
1369     } while ((index < htabslots) && !qemu_file_rate_limit(f));
1370 
1371     if (index >= htabslots) {
1372         assert(index == htabslots);
1373         index = 0;
1374         spapr->htab_first_pass = false;
1375     }
1376     spapr->htab_save_index = index;
1377 }
1378 
1379 static int htab_save_later_pass(QEMUFile *f, sPAPRMachineState *spapr,
1380                                 int64_t max_ns)
1381 {
1382     bool final = max_ns < 0;
1383     int htabslots = HTAB_SIZE(spapr) / HASH_PTE_SIZE_64;
1384     int examined = 0, sent = 0;
1385     int index = spapr->htab_save_index;
1386     int64_t starttime = qemu_clock_get_ns(QEMU_CLOCK_REALTIME);
1387 
1388     assert(!spapr->htab_first_pass);
1389 
1390     do {
1391         int chunkstart, invalidstart;
1392 
1393         /* Consume non-dirty HPTEs */
1394         while ((index < htabslots)
1395                && !HPTE_DIRTY(HPTE(spapr->htab, index))) {
1396             index++;
1397             examined++;
1398         }
1399 
1400         chunkstart = index;
1401         /* Consume valid dirty HPTEs */
1402         while ((index < htabslots) && (index - chunkstart < USHRT_MAX)
1403                && HPTE_DIRTY(HPTE(spapr->htab, index))
1404                && HPTE_VALID(HPTE(spapr->htab, index))) {
1405             CLEAN_HPTE(HPTE(spapr->htab, index));
1406             index++;
1407             examined++;
1408         }
1409 
1410         invalidstart = index;
1411         /* Consume invalid dirty HPTEs */
1412         while ((index < htabslots) && (index - invalidstart < USHRT_MAX)
1413                && HPTE_DIRTY(HPTE(spapr->htab, index))
1414                && !HPTE_VALID(HPTE(spapr->htab, index))) {
1415             CLEAN_HPTE(HPTE(spapr->htab, index));
1416             index++;
1417             examined++;
1418         }
1419 
1420         if (index > chunkstart) {
1421             int n_valid = invalidstart - chunkstart;
1422             int n_invalid = index - invalidstart;
1423 
1424             qemu_put_be32(f, chunkstart);
1425             qemu_put_be16(f, n_valid);
1426             qemu_put_be16(f, n_invalid);
1427             qemu_put_buffer(f, HPTE(spapr->htab, chunkstart),
1428                             HASH_PTE_SIZE_64 * n_valid);
1429             sent += index - chunkstart;
1430 
1431             if (!final && (qemu_clock_get_ns(QEMU_CLOCK_REALTIME) - starttime) > max_ns) {
1432                 break;
1433             }
1434         }
1435 
1436         if (examined >= htabslots) {
1437             break;
1438         }
1439 
1440         if (index >= htabslots) {
1441             assert(index == htabslots);
1442             index = 0;
1443         }
1444     } while ((examined < htabslots) && (!qemu_file_rate_limit(f) || final));
1445 
1446     if (index >= htabslots) {
1447         assert(index == htabslots);
1448         index = 0;
1449     }
1450 
1451     spapr->htab_save_index = index;
1452 
1453     return (examined >= htabslots) && (sent == 0) ? 1 : 0;
1454 }
1455 
1456 #define MAX_ITERATION_NS    5000000 /* 5 ms */
1457 #define MAX_KVM_BUF_SIZE    2048
1458 
1459 static int htab_save_iterate(QEMUFile *f, void *opaque)
1460 {
1461     sPAPRMachineState *spapr = opaque;
1462     int fd;
1463     int rc = 0;
1464 
1465     /* Iteration header */
1466     qemu_put_be32(f, 0);
1467 
1468     if (!spapr->htab) {
1469         assert(kvm_enabled());
1470 
1471         fd = get_htab_fd(spapr);
1472         if (fd < 0) {
1473             return fd;
1474         }
1475 
1476         rc = kvmppc_save_htab(f, fd, MAX_KVM_BUF_SIZE, MAX_ITERATION_NS);
1477         if (rc < 0) {
1478             return rc;
1479         }
1480     } else  if (spapr->htab_first_pass) {
1481         htab_save_first_pass(f, spapr, MAX_ITERATION_NS);
1482     } else {
1483         rc = htab_save_later_pass(f, spapr, MAX_ITERATION_NS);
1484     }
1485 
1486     /* End marker */
1487     qemu_put_be32(f, 0);
1488     qemu_put_be16(f, 0);
1489     qemu_put_be16(f, 0);
1490 
1491     return rc;
1492 }
1493 
1494 static int htab_save_complete(QEMUFile *f, void *opaque)
1495 {
1496     sPAPRMachineState *spapr = opaque;
1497     int fd;
1498 
1499     /* Iteration header */
1500     qemu_put_be32(f, 0);
1501 
1502     if (!spapr->htab) {
1503         int rc;
1504 
1505         assert(kvm_enabled());
1506 
1507         fd = get_htab_fd(spapr);
1508         if (fd < 0) {
1509             return fd;
1510         }
1511 
1512         rc = kvmppc_save_htab(f, fd, MAX_KVM_BUF_SIZE, -1);
1513         if (rc < 0) {
1514             return rc;
1515         }
1516     } else {
1517         if (spapr->htab_first_pass) {
1518             htab_save_first_pass(f, spapr, -1);
1519         }
1520         htab_save_later_pass(f, spapr, -1);
1521     }
1522 
1523     /* End marker */
1524     qemu_put_be32(f, 0);
1525     qemu_put_be16(f, 0);
1526     qemu_put_be16(f, 0);
1527 
1528     return 0;
1529 }
1530 
1531 static int htab_load(QEMUFile *f, void *opaque, int version_id)
1532 {
1533     sPAPRMachineState *spapr = opaque;
1534     uint32_t section_hdr;
1535     int fd = -1;
1536 
1537     if (version_id < 1 || version_id > 1) {
1538         error_report("htab_load() bad version");
1539         return -EINVAL;
1540     }
1541 
1542     section_hdr = qemu_get_be32(f);
1543 
1544     if (section_hdr) {
1545         Error *local_err = NULL;
1546 
1547         /* First section gives the htab size */
1548         spapr_reallocate_hpt(spapr, section_hdr, &local_err);
1549         if (local_err) {
1550             error_report_err(local_err);
1551             return -EINVAL;
1552         }
1553         return 0;
1554     }
1555 
1556     if (!spapr->htab) {
1557         assert(kvm_enabled());
1558 
1559         fd = kvmppc_get_htab_fd(true);
1560         if (fd < 0) {
1561             error_report("Unable to open fd to restore KVM hash table: %s",
1562                          strerror(errno));
1563         }
1564     }
1565 
1566     while (true) {
1567         uint32_t index;
1568         uint16_t n_valid, n_invalid;
1569 
1570         index = qemu_get_be32(f);
1571         n_valid = qemu_get_be16(f);
1572         n_invalid = qemu_get_be16(f);
1573 
1574         if ((index == 0) && (n_valid == 0) && (n_invalid == 0)) {
1575             /* End of Stream */
1576             break;
1577         }
1578 
1579         if ((index + n_valid + n_invalid) >
1580             (HTAB_SIZE(spapr) / HASH_PTE_SIZE_64)) {
1581             /* Bad index in stream */
1582             error_report(
1583                 "htab_load() bad index %d (%hd+%hd entries) in htab stream (htab_shift=%d)",
1584                 index, n_valid, n_invalid, spapr->htab_shift);
1585             return -EINVAL;
1586         }
1587 
1588         if (spapr->htab) {
1589             if (n_valid) {
1590                 qemu_get_buffer(f, HPTE(spapr->htab, index),
1591                                 HASH_PTE_SIZE_64 * n_valid);
1592             }
1593             if (n_invalid) {
1594                 memset(HPTE(spapr->htab, index + n_valid), 0,
1595                        HASH_PTE_SIZE_64 * n_invalid);
1596             }
1597         } else {
1598             int rc;
1599 
1600             assert(fd >= 0);
1601 
1602             rc = kvmppc_load_htab_chunk(f, fd, index, n_valid, n_invalid);
1603             if (rc < 0) {
1604                 return rc;
1605             }
1606         }
1607     }
1608 
1609     if (!spapr->htab) {
1610         assert(fd >= 0);
1611         close(fd);
1612     }
1613 
1614     return 0;
1615 }
1616 
1617 static void htab_cleanup(void *opaque)
1618 {
1619     sPAPRMachineState *spapr = opaque;
1620 
1621     close_htab_fd(spapr);
1622 }
1623 
1624 static SaveVMHandlers savevm_htab_handlers = {
1625     .save_live_setup = htab_save_setup,
1626     .save_live_iterate = htab_save_iterate,
1627     .save_live_complete_precopy = htab_save_complete,
1628     .cleanup = htab_cleanup,
1629     .load_state = htab_load,
1630 };
1631 
1632 static void spapr_boot_set(void *opaque, const char *boot_device,
1633                            Error **errp)
1634 {
1635     MachineState *machine = MACHINE(qdev_get_machine());
1636     machine->boot_order = g_strdup(boot_device);
1637 }
1638 
1639 /*
1640  * Reset routine for LMB DR devices.
1641  *
1642  * Unlike PCI DR devices, LMB DR devices explicitly register this reset
1643  * routine. Reset for PCI DR devices will be handled by PHB reset routine
1644  * when it walks all its children devices. LMB devices reset occurs
1645  * as part of spapr_ppc_reset().
1646  */
1647 static void spapr_drc_reset(void *opaque)
1648 {
1649     sPAPRDRConnector *drc = opaque;
1650     DeviceState *d = DEVICE(drc);
1651 
1652     if (d) {
1653         device_reset(d);
1654     }
1655 }
1656 
1657 static void spapr_create_lmb_dr_connectors(sPAPRMachineState *spapr)
1658 {
1659     MachineState *machine = MACHINE(spapr);
1660     uint64_t lmb_size = SPAPR_MEMORY_BLOCK_SIZE;
1661     uint32_t nr_lmbs = (machine->maxram_size - machine->ram_size)/lmb_size;
1662     int i;
1663 
1664     for (i = 0; i < nr_lmbs; i++) {
1665         sPAPRDRConnector *drc;
1666         uint64_t addr;
1667 
1668         addr = i * lmb_size + spapr->hotplug_memory.base;
1669         drc = spapr_dr_connector_new(OBJECT(spapr), SPAPR_DR_CONNECTOR_TYPE_LMB,
1670                                      addr/lmb_size);
1671         qemu_register_reset(spapr_drc_reset, drc);
1672     }
1673 }
1674 
1675 /*
1676  * If RAM size, maxmem size and individual node mem sizes aren't aligned
1677  * to SPAPR_MEMORY_BLOCK_SIZE(256MB), then refuse to start the guest
1678  * since we can't support such unaligned sizes with DRCONF_MEMORY.
1679  */
1680 static void spapr_validate_node_memory(MachineState *machine, Error **errp)
1681 {
1682     int i;
1683 
1684     if (machine->ram_size % SPAPR_MEMORY_BLOCK_SIZE) {
1685         error_setg(errp, "Memory size 0x" RAM_ADDR_FMT
1686                    " is not aligned to %llu MiB",
1687                    machine->ram_size,
1688                    SPAPR_MEMORY_BLOCK_SIZE / M_BYTE);
1689         return;
1690     }
1691 
1692     if (machine->maxram_size % SPAPR_MEMORY_BLOCK_SIZE) {
1693         error_setg(errp, "Maximum memory size 0x" RAM_ADDR_FMT
1694                    " is not aligned to %llu MiB",
1695                    machine->ram_size,
1696                    SPAPR_MEMORY_BLOCK_SIZE / M_BYTE);
1697         return;
1698     }
1699 
1700     for (i = 0; i < nb_numa_nodes; i++) {
1701         if (numa_info[i].node_mem % SPAPR_MEMORY_BLOCK_SIZE) {
1702             error_setg(errp,
1703                        "Node %d memory size 0x%" PRIx64
1704                        " is not aligned to %llu MiB",
1705                        i, numa_info[i].node_mem,
1706                        SPAPR_MEMORY_BLOCK_SIZE / M_BYTE);
1707             return;
1708         }
1709     }
1710 }
1711 
1712 /* pSeries LPAR / sPAPR hardware init */
1713 static void ppc_spapr_init(MachineState *machine)
1714 {
1715     sPAPRMachineState *spapr = SPAPR_MACHINE(machine);
1716     MachineClass *mc = MACHINE_GET_CLASS(machine);
1717     sPAPRMachineClass *smc = SPAPR_MACHINE_GET_CLASS(machine);
1718     const char *kernel_filename = machine->kernel_filename;
1719     const char *kernel_cmdline = machine->kernel_cmdline;
1720     const char *initrd_filename = machine->initrd_filename;
1721     PCIHostState *phb;
1722     int i;
1723     MemoryRegion *sysmem = get_system_memory();
1724     MemoryRegion *ram = g_new(MemoryRegion, 1);
1725     MemoryRegion *rma_region;
1726     void *rma = NULL;
1727     hwaddr rma_alloc_size;
1728     hwaddr node0_size = spapr_node0_size();
1729     uint32_t initrd_base = 0;
1730     long kernel_size = 0, initrd_size = 0;
1731     long load_limit, fw_size;
1732     bool kernel_le = false;
1733     char *filename;
1734     int smt = kvmppc_smt_threads();
1735     int spapr_cores = smp_cpus / smp_threads;
1736     int spapr_max_cores = max_cpus / smp_threads;
1737 
1738     if (mc->query_hotpluggable_cpus) {
1739         if (smp_cpus % smp_threads) {
1740             error_report("smp_cpus (%u) must be multiple of threads (%u)",
1741                          smp_cpus, smp_threads);
1742             exit(1);
1743         }
1744         if (max_cpus % smp_threads) {
1745             error_report("max_cpus (%u) must be multiple of threads (%u)",
1746                          max_cpus, smp_threads);
1747             exit(1);
1748         }
1749     }
1750 
1751     msi_nonbroken = true;
1752 
1753     QLIST_INIT(&spapr->phbs);
1754 
1755     cpu_ppc_hypercall = emulate_spapr_hypercall;
1756 
1757     /* Allocate RMA if necessary */
1758     rma_alloc_size = kvmppc_alloc_rma(&rma);
1759 
1760     if (rma_alloc_size == -1) {
1761         error_report("Unable to create RMA");
1762         exit(1);
1763     }
1764 
1765     if (rma_alloc_size && (rma_alloc_size < node0_size)) {
1766         spapr->rma_size = rma_alloc_size;
1767     } else {
1768         spapr->rma_size = node0_size;
1769 
1770         /* With KVM, we don't actually know whether KVM supports an
1771          * unbounded RMA (PR KVM) or is limited by the hash table size
1772          * (HV KVM using VRMA), so we always assume the latter
1773          *
1774          * In that case, we also limit the initial allocations for RTAS
1775          * etc... to 256M since we have no way to know what the VRMA size
1776          * is going to be as it depends on the size of the hash table
1777          * isn't determined yet.
1778          */
1779         if (kvm_enabled()) {
1780             spapr->vrma_adjust = 1;
1781             spapr->rma_size = MIN(spapr->rma_size, 0x10000000);
1782         }
1783 
1784         /* Actually we don't support unbounded RMA anymore since we
1785          * added proper emulation of HV mode. The max we can get is
1786          * 16G which also happens to be what we configure for PAPR
1787          * mode so make sure we don't do anything bigger than that
1788          */
1789         spapr->rma_size = MIN(spapr->rma_size, 0x400000000ull);
1790     }
1791 
1792     if (spapr->rma_size > node0_size) {
1793         error_report("Numa node 0 has to span the RMA (%#08"HWADDR_PRIx")",
1794                      spapr->rma_size);
1795         exit(1);
1796     }
1797 
1798     /* Setup a load limit for the ramdisk leaving room for SLOF and FDT */
1799     load_limit = MIN(spapr->rma_size, RTAS_MAX_ADDR) - FW_OVERHEAD;
1800 
1801     /* Set up Interrupt Controller before we create the VCPUs */
1802     spapr->xics = xics_system_init(machine,
1803                                    DIV_ROUND_UP(max_cpus * smt, smp_threads),
1804                                    XICS_IRQS_SPAPR, &error_fatal);
1805 
1806     if (smc->dr_lmb_enabled) {
1807         spapr_validate_node_memory(machine, &error_fatal);
1808     }
1809 
1810     /* init CPUs */
1811     if (machine->cpu_model == NULL) {
1812         machine->cpu_model = kvm_enabled() ? "host" : "POWER7";
1813     }
1814 
1815     if (mc->query_hotpluggable_cpus) {
1816         char *type = spapr_get_cpu_core_type(machine->cpu_model);
1817 
1818         spapr->cores = g_new0(Object *, spapr_max_cores);
1819         for (i = 0; i < spapr_max_cores; i++) {
1820             int core_id = i * smp_threads;
1821             sPAPRDRConnector *drc =
1822                 spapr_dr_connector_new(OBJECT(spapr),
1823                                        SPAPR_DR_CONNECTOR_TYPE_CPU,
1824                                        (core_id / smp_threads) * smt);
1825 
1826             qemu_register_reset(spapr_drc_reset, drc);
1827 
1828             if (i < spapr_cores) {
1829                 char *type = spapr_get_cpu_core_type(machine->cpu_model);
1830                 Object *core;
1831 
1832                 if (!object_class_by_name(type)) {
1833                     error_report("Unable to find sPAPR CPU Core definition");
1834                     exit(1);
1835                 }
1836 
1837                 core  = object_new(type);
1838                 object_property_set_int(core, smp_threads, "nr-threads",
1839                                         &error_fatal);
1840                 object_property_set_int(core, core_id, CPU_CORE_PROP_CORE_ID,
1841                                         &error_fatal);
1842                 object_property_set_bool(core, true, "realized", &error_fatal);
1843             }
1844         }
1845         g_free(type);
1846     } else {
1847         for (i = 0; i < smp_cpus; i++) {
1848             PowerPCCPU *cpu = cpu_ppc_init(machine->cpu_model);
1849             if (cpu == NULL) {
1850                 error_report("Unable to find PowerPC CPU definition");
1851                 exit(1);
1852             }
1853             spapr_cpu_init(spapr, cpu, &error_fatal);
1854        }
1855     }
1856 
1857     if (kvm_enabled()) {
1858         /* Enable H_LOGICAL_CI_* so SLOF can talk to in-kernel devices */
1859         kvmppc_enable_logical_ci_hcalls();
1860         kvmppc_enable_set_mode_hcall();
1861     }
1862 
1863     /* allocate RAM */
1864     memory_region_allocate_system_memory(ram, NULL, "ppc_spapr.ram",
1865                                          machine->ram_size);
1866     memory_region_add_subregion(sysmem, 0, ram);
1867 
1868     if (rma_alloc_size && rma) {
1869         rma_region = g_new(MemoryRegion, 1);
1870         memory_region_init_ram_ptr(rma_region, NULL, "ppc_spapr.rma",
1871                                    rma_alloc_size, rma);
1872         vmstate_register_ram_global(rma_region);
1873         memory_region_add_subregion(sysmem, 0, rma_region);
1874     }
1875 
1876     /* initialize hotplug memory address space */
1877     if (machine->ram_size < machine->maxram_size) {
1878         ram_addr_t hotplug_mem_size = machine->maxram_size - machine->ram_size;
1879         /*
1880          * Limit the number of hotpluggable memory slots to half the number
1881          * slots that KVM supports, leaving the other half for PCI and other
1882          * devices. However ensure that number of slots doesn't drop below 32.
1883          */
1884         int max_memslots = kvm_enabled() ? kvm_get_max_memslots() / 2 :
1885                            SPAPR_MAX_RAM_SLOTS;
1886 
1887         if (max_memslots < SPAPR_MAX_RAM_SLOTS) {
1888             max_memslots = SPAPR_MAX_RAM_SLOTS;
1889         }
1890         if (machine->ram_slots > max_memslots) {
1891             error_report("Specified number of memory slots %"
1892                          PRIu64" exceeds max supported %d",
1893                          machine->ram_slots, max_memslots);
1894             exit(1);
1895         }
1896 
1897         spapr->hotplug_memory.base = ROUND_UP(machine->ram_size,
1898                                               SPAPR_HOTPLUG_MEM_ALIGN);
1899         memory_region_init(&spapr->hotplug_memory.mr, OBJECT(spapr),
1900                            "hotplug-memory", hotplug_mem_size);
1901         memory_region_add_subregion(sysmem, spapr->hotplug_memory.base,
1902                                     &spapr->hotplug_memory.mr);
1903     }
1904 
1905     if (smc->dr_lmb_enabled) {
1906         spapr_create_lmb_dr_connectors(spapr);
1907     }
1908 
1909     filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, "spapr-rtas.bin");
1910     if (!filename) {
1911         error_report("Could not find LPAR rtas '%s'", "spapr-rtas.bin");
1912         exit(1);
1913     }
1914     spapr->rtas_size = get_image_size(filename);
1915     if (spapr->rtas_size < 0) {
1916         error_report("Could not get size of LPAR rtas '%s'", filename);
1917         exit(1);
1918     }
1919     spapr->rtas_blob = g_malloc(spapr->rtas_size);
1920     if (load_image_size(filename, spapr->rtas_blob, spapr->rtas_size) < 0) {
1921         error_report("Could not load LPAR rtas '%s'", filename);
1922         exit(1);
1923     }
1924     if (spapr->rtas_size > RTAS_MAX_SIZE) {
1925         error_report("RTAS too big ! 0x%zx bytes (max is 0x%x)",
1926                      (size_t)spapr->rtas_size, RTAS_MAX_SIZE);
1927         exit(1);
1928     }
1929     g_free(filename);
1930 
1931     /* Set up EPOW events infrastructure */
1932     spapr_events_init(spapr);
1933 
1934     /* Set up the RTC RTAS interfaces */
1935     spapr_rtc_create(spapr);
1936 
1937     /* Set up VIO bus */
1938     spapr->vio_bus = spapr_vio_bus_init();
1939 
1940     for (i = 0; i < MAX_SERIAL_PORTS; i++) {
1941         if (serial_hds[i]) {
1942             spapr_vty_create(spapr->vio_bus, serial_hds[i]);
1943         }
1944     }
1945 
1946     /* We always have at least the nvram device on VIO */
1947     spapr_create_nvram(spapr);
1948 
1949     /* Set up PCI */
1950     spapr_pci_rtas_init();
1951 
1952     phb = spapr_create_phb(spapr, 0);
1953 
1954     for (i = 0; i < nb_nics; i++) {
1955         NICInfo *nd = &nd_table[i];
1956 
1957         if (!nd->model) {
1958             nd->model = g_strdup("ibmveth");
1959         }
1960 
1961         if (strcmp(nd->model, "ibmveth") == 0) {
1962             spapr_vlan_create(spapr->vio_bus, nd);
1963         } else {
1964             pci_nic_init_nofail(&nd_table[i], phb->bus, nd->model, NULL);
1965         }
1966     }
1967 
1968     for (i = 0; i <= drive_get_max_bus(IF_SCSI); i++) {
1969         spapr_vscsi_create(spapr->vio_bus);
1970     }
1971 
1972     /* Graphics */
1973     if (spapr_vga_init(phb->bus, &error_fatal)) {
1974         spapr->has_graphics = true;
1975         machine->usb |= defaults_enabled() && !machine->usb_disabled;
1976     }
1977 
1978     if (machine->usb) {
1979         if (smc->use_ohci_by_default) {
1980             pci_create_simple(phb->bus, -1, "pci-ohci");
1981         } else {
1982             pci_create_simple(phb->bus, -1, "nec-usb-xhci");
1983         }
1984 
1985         if (spapr->has_graphics) {
1986             USBBus *usb_bus = usb_bus_find(-1);
1987 
1988             usb_create_simple(usb_bus, "usb-kbd");
1989             usb_create_simple(usb_bus, "usb-mouse");
1990         }
1991     }
1992 
1993     if (spapr->rma_size < (MIN_RMA_SLOF << 20)) {
1994         error_report(
1995             "pSeries SLOF firmware requires >= %ldM guest RMA (Real Mode Area memory)",
1996             MIN_RMA_SLOF);
1997         exit(1);
1998     }
1999 
2000     if (kernel_filename) {
2001         uint64_t lowaddr = 0;
2002 
2003         kernel_size = load_elf(kernel_filename, translate_kernel_address, NULL,
2004                                NULL, &lowaddr, NULL, 1, PPC_ELF_MACHINE,
2005                                0, 0);
2006         if (kernel_size == ELF_LOAD_WRONG_ENDIAN) {
2007             kernel_size = load_elf(kernel_filename,
2008                                    translate_kernel_address, NULL,
2009                                    NULL, &lowaddr, NULL, 0, PPC_ELF_MACHINE,
2010                                    0, 0);
2011             kernel_le = kernel_size > 0;
2012         }
2013         if (kernel_size < 0) {
2014             error_report("error loading %s: %s",
2015                          kernel_filename, load_elf_strerror(kernel_size));
2016             exit(1);
2017         }
2018 
2019         /* load initrd */
2020         if (initrd_filename) {
2021             /* Try to locate the initrd in the gap between the kernel
2022              * and the firmware. Add a bit of space just in case
2023              */
2024             initrd_base = (KERNEL_LOAD_ADDR + kernel_size + 0x1ffff) & ~0xffff;
2025             initrd_size = load_image_targphys(initrd_filename, initrd_base,
2026                                               load_limit - initrd_base);
2027             if (initrd_size < 0) {
2028                 error_report("could not load initial ram disk '%s'",
2029                              initrd_filename);
2030                 exit(1);
2031             }
2032         } else {
2033             initrd_base = 0;
2034             initrd_size = 0;
2035         }
2036     }
2037 
2038     if (bios_name == NULL) {
2039         bios_name = FW_FILE_NAME;
2040     }
2041     filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, bios_name);
2042     if (!filename) {
2043         error_report("Could not find LPAR firmware '%s'", bios_name);
2044         exit(1);
2045     }
2046     fw_size = load_image_targphys(filename, 0, FW_MAX_SIZE);
2047     if (fw_size <= 0) {
2048         error_report("Could not load LPAR firmware '%s'", filename);
2049         exit(1);
2050     }
2051     g_free(filename);
2052 
2053     /* FIXME: Should register things through the MachineState's qdev
2054      * interface, this is a legacy from the sPAPREnvironment structure
2055      * which predated MachineState but had a similar function */
2056     vmstate_register(NULL, 0, &vmstate_spapr, spapr);
2057     register_savevm_live(NULL, "spapr/htab", -1, 1,
2058                          &savevm_htab_handlers, spapr);
2059 
2060     /* Prepare the device tree */
2061     spapr->fdt_skel = spapr_create_fdt_skel(initrd_base, initrd_size,
2062                                             kernel_size, kernel_le,
2063                                             kernel_cmdline,
2064                                             spapr->check_exception_irq);
2065     assert(spapr->fdt_skel != NULL);
2066 
2067     /* used by RTAS */
2068     QTAILQ_INIT(&spapr->ccs_list);
2069     qemu_register_reset(spapr_ccs_reset_hook, spapr);
2070 
2071     qemu_register_boot_set(spapr_boot_set, spapr);
2072 }
2073 
2074 static int spapr_kvm_type(const char *vm_type)
2075 {
2076     if (!vm_type) {
2077         return 0;
2078     }
2079 
2080     if (!strcmp(vm_type, "HV")) {
2081         return 1;
2082     }
2083 
2084     if (!strcmp(vm_type, "PR")) {
2085         return 2;
2086     }
2087 
2088     error_report("Unknown kvm-type specified '%s'", vm_type);
2089     exit(1);
2090 }
2091 
2092 /*
2093  * Implementation of an interface to adjust firmware path
2094  * for the bootindex property handling.
2095  */
2096 static char *spapr_get_fw_dev_path(FWPathProvider *p, BusState *bus,
2097                                    DeviceState *dev)
2098 {
2099 #define CAST(type, obj, name) \
2100     ((type *)object_dynamic_cast(OBJECT(obj), (name)))
2101     SCSIDevice *d = CAST(SCSIDevice,  dev, TYPE_SCSI_DEVICE);
2102     sPAPRPHBState *phb = CAST(sPAPRPHBState, dev, TYPE_SPAPR_PCI_HOST_BRIDGE);
2103 
2104     if (d) {
2105         void *spapr = CAST(void, bus->parent, "spapr-vscsi");
2106         VirtIOSCSI *virtio = CAST(VirtIOSCSI, bus->parent, TYPE_VIRTIO_SCSI);
2107         USBDevice *usb = CAST(USBDevice, bus->parent, TYPE_USB_DEVICE);
2108 
2109         if (spapr) {
2110             /*
2111              * Replace "channel@0/disk@0,0" with "disk@8000000000000000":
2112              * We use SRP luns of the form 8000 | (bus << 8) | (id << 5) | lun
2113              * in the top 16 bits of the 64-bit LUN
2114              */
2115             unsigned id = 0x8000 | (d->id << 8) | d->lun;
2116             return g_strdup_printf("%s@%"PRIX64, qdev_fw_name(dev),
2117                                    (uint64_t)id << 48);
2118         } else if (virtio) {
2119             /*
2120              * We use SRP luns of the form 01000000 | (target << 8) | lun
2121              * in the top 32 bits of the 64-bit LUN
2122              * Note: the quote above is from SLOF and it is wrong,
2123              * the actual binding is:
2124              * swap 0100 or 10 << or 20 << ( target lun-id -- srplun )
2125              */
2126             unsigned id = 0x1000000 | (d->id << 16) | d->lun;
2127             return g_strdup_printf("%s@%"PRIX64, qdev_fw_name(dev),
2128                                    (uint64_t)id << 32);
2129         } else if (usb) {
2130             /*
2131              * We use SRP luns of the form 01000000 | (usb-port << 16) | lun
2132              * in the top 32 bits of the 64-bit LUN
2133              */
2134             unsigned usb_port = atoi(usb->port->path);
2135             unsigned id = 0x1000000 | (usb_port << 16) | d->lun;
2136             return g_strdup_printf("%s@%"PRIX64, qdev_fw_name(dev),
2137                                    (uint64_t)id << 32);
2138         }
2139     }
2140 
2141     if (phb) {
2142         /* Replace "pci" with "pci@800000020000000" */
2143         return g_strdup_printf("pci@%"PRIX64, phb->buid);
2144     }
2145 
2146     return NULL;
2147 }
2148 
2149 static char *spapr_get_kvm_type(Object *obj, Error **errp)
2150 {
2151     sPAPRMachineState *spapr = SPAPR_MACHINE(obj);
2152 
2153     return g_strdup(spapr->kvm_type);
2154 }
2155 
2156 static void spapr_set_kvm_type(Object *obj, const char *value, Error **errp)
2157 {
2158     sPAPRMachineState *spapr = SPAPR_MACHINE(obj);
2159 
2160     g_free(spapr->kvm_type);
2161     spapr->kvm_type = g_strdup(value);
2162 }
2163 
2164 static void spapr_machine_initfn(Object *obj)
2165 {
2166     sPAPRMachineState *spapr = SPAPR_MACHINE(obj);
2167 
2168     spapr->htab_fd = -1;
2169     object_property_add_str(obj, "kvm-type",
2170                             spapr_get_kvm_type, spapr_set_kvm_type, NULL);
2171     object_property_set_description(obj, "kvm-type",
2172                                     "Specifies the KVM virtualization mode (HV, PR)",
2173                                     NULL);
2174 }
2175 
2176 static void spapr_machine_finalizefn(Object *obj)
2177 {
2178     sPAPRMachineState *spapr = SPAPR_MACHINE(obj);
2179 
2180     g_free(spapr->kvm_type);
2181 }
2182 
2183 static void ppc_cpu_do_nmi_on_cpu(void *arg)
2184 {
2185     CPUState *cs = arg;
2186 
2187     cpu_synchronize_state(cs);
2188     ppc_cpu_do_system_reset(cs);
2189 }
2190 
2191 static void spapr_nmi(NMIState *n, int cpu_index, Error **errp)
2192 {
2193     CPUState *cs;
2194 
2195     CPU_FOREACH(cs) {
2196         async_run_on_cpu(cs, ppc_cpu_do_nmi_on_cpu, cs);
2197     }
2198 }
2199 
2200 static void spapr_add_lmbs(DeviceState *dev, uint64_t addr, uint64_t size,
2201                            uint32_t node, Error **errp)
2202 {
2203     sPAPRDRConnector *drc;
2204     sPAPRDRConnectorClass *drck;
2205     uint32_t nr_lmbs = size/SPAPR_MEMORY_BLOCK_SIZE;
2206     int i, fdt_offset, fdt_size;
2207     void *fdt;
2208 
2209     for (i = 0; i < nr_lmbs; i++) {
2210         drc = spapr_dr_connector_by_id(SPAPR_DR_CONNECTOR_TYPE_LMB,
2211                 addr/SPAPR_MEMORY_BLOCK_SIZE);
2212         g_assert(drc);
2213 
2214         fdt = create_device_tree(&fdt_size);
2215         fdt_offset = spapr_populate_memory_node(fdt, node, addr,
2216                                                 SPAPR_MEMORY_BLOCK_SIZE);
2217 
2218         drck = SPAPR_DR_CONNECTOR_GET_CLASS(drc);
2219         drck->attach(drc, dev, fdt, fdt_offset, !dev->hotplugged, errp);
2220         addr += SPAPR_MEMORY_BLOCK_SIZE;
2221     }
2222     /* send hotplug notification to the
2223      * guest only in case of hotplugged memory
2224      */
2225     if (dev->hotplugged) {
2226        spapr_hotplug_req_add_by_count(SPAPR_DR_CONNECTOR_TYPE_LMB, nr_lmbs);
2227     }
2228 }
2229 
2230 static void spapr_memory_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
2231                               uint32_t node, Error **errp)
2232 {
2233     Error *local_err = NULL;
2234     sPAPRMachineState *ms = SPAPR_MACHINE(hotplug_dev);
2235     PCDIMMDevice *dimm = PC_DIMM(dev);
2236     PCDIMMDeviceClass *ddc = PC_DIMM_GET_CLASS(dimm);
2237     MemoryRegion *mr = ddc->get_memory_region(dimm);
2238     uint64_t align = memory_region_get_alignment(mr);
2239     uint64_t size = memory_region_size(mr);
2240     uint64_t addr;
2241 
2242     if (size % SPAPR_MEMORY_BLOCK_SIZE) {
2243         error_setg(&local_err, "Hotplugged memory size must be a multiple of "
2244                       "%lld MB", SPAPR_MEMORY_BLOCK_SIZE/M_BYTE);
2245         goto out;
2246     }
2247 
2248     pc_dimm_memory_plug(dev, &ms->hotplug_memory, mr, align, &local_err);
2249     if (local_err) {
2250         goto out;
2251     }
2252 
2253     addr = object_property_get_int(OBJECT(dimm), PC_DIMM_ADDR_PROP, &local_err);
2254     if (local_err) {
2255         pc_dimm_memory_unplug(dev, &ms->hotplug_memory, mr);
2256         goto out;
2257     }
2258 
2259     spapr_add_lmbs(dev, addr, size, node, &error_abort);
2260 
2261 out:
2262     error_propagate(errp, local_err);
2263 }
2264 
2265 void *spapr_populate_hotplug_cpu_dt(CPUState *cs, int *fdt_offset,
2266                                     sPAPRMachineState *spapr)
2267 {
2268     PowerPCCPU *cpu = POWERPC_CPU(cs);
2269     DeviceClass *dc = DEVICE_GET_CLASS(cs);
2270     int id = ppc_get_vcpu_dt_id(cpu);
2271     void *fdt;
2272     int offset, fdt_size;
2273     char *nodename;
2274 
2275     fdt = create_device_tree(&fdt_size);
2276     nodename = g_strdup_printf("%s@%x", dc->fw_name, id);
2277     offset = fdt_add_subnode(fdt, 0, nodename);
2278 
2279     spapr_populate_cpu_dt(cs, fdt, offset, spapr);
2280     g_free(nodename);
2281 
2282     *fdt_offset = offset;
2283     return fdt;
2284 }
2285 
2286 static void spapr_machine_device_plug(HotplugHandler *hotplug_dev,
2287                                       DeviceState *dev, Error **errp)
2288 {
2289     sPAPRMachineClass *smc = SPAPR_MACHINE_GET_CLASS(qdev_get_machine());
2290 
2291     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2292         int node;
2293 
2294         if (!smc->dr_lmb_enabled) {
2295             error_setg(errp, "Memory hotplug not supported for this machine");
2296             return;
2297         }
2298         node = object_property_get_int(OBJECT(dev), PC_DIMM_NODE_PROP, errp);
2299         if (*errp) {
2300             return;
2301         }
2302         if (node < 0 || node >= MAX_NODES) {
2303             error_setg(errp, "Invaild node %d", node);
2304             return;
2305         }
2306 
2307         /*
2308          * Currently PowerPC kernel doesn't allow hot-adding memory to
2309          * memory-less node, but instead will silently add the memory
2310          * to the first node that has some memory. This causes two
2311          * unexpected behaviours for the user.
2312          *
2313          * - Memory gets hotplugged to a different node than what the user
2314          *   specified.
2315          * - Since pc-dimm subsystem in QEMU still thinks that memory belongs
2316          *   to memory-less node, a reboot will set things accordingly
2317          *   and the previously hotplugged memory now ends in the right node.
2318          *   This appears as if some memory moved from one node to another.
2319          *
2320          * So until kernel starts supporting memory hotplug to memory-less
2321          * nodes, just prevent such attempts upfront in QEMU.
2322          */
2323         if (nb_numa_nodes && !numa_info[node].node_mem) {
2324             error_setg(errp, "Can't hotplug memory to memory-less node %d",
2325                        node);
2326             return;
2327         }
2328 
2329         spapr_memory_plug(hotplug_dev, dev, node, errp);
2330     } else if (object_dynamic_cast(OBJECT(dev), TYPE_SPAPR_CPU_CORE)) {
2331         spapr_core_plug(hotplug_dev, dev, errp);
2332     }
2333 }
2334 
2335 static void spapr_machine_device_unplug(HotplugHandler *hotplug_dev,
2336                                       DeviceState *dev, Error **errp)
2337 {
2338     MachineClass *mc = MACHINE_GET_CLASS(qdev_get_machine());
2339 
2340     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2341         error_setg(errp, "Memory hot unplug not supported by sPAPR");
2342     } else if (object_dynamic_cast(OBJECT(dev), TYPE_SPAPR_CPU_CORE)) {
2343         if (!mc->query_hotpluggable_cpus) {
2344             error_setg(errp, "CPU hot unplug not supported on this machine");
2345             return;
2346         }
2347         spapr_core_unplug(hotplug_dev, dev, errp);
2348     }
2349 }
2350 
2351 static void spapr_machine_device_pre_plug(HotplugHandler *hotplug_dev,
2352                                           DeviceState *dev, Error **errp)
2353 {
2354     if (object_dynamic_cast(OBJECT(dev), TYPE_SPAPR_CPU_CORE)) {
2355         spapr_core_pre_plug(hotplug_dev, dev, errp);
2356     }
2357 }
2358 
2359 static HotplugHandler *spapr_get_hotpug_handler(MachineState *machine,
2360                                              DeviceState *dev)
2361 {
2362     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) ||
2363         object_dynamic_cast(OBJECT(dev), TYPE_SPAPR_CPU_CORE)) {
2364         return HOTPLUG_HANDLER(machine);
2365     }
2366     return NULL;
2367 }
2368 
2369 static unsigned spapr_cpu_index_to_socket_id(unsigned cpu_index)
2370 {
2371     /* Allocate to NUMA nodes on a "socket" basis (not that concept of
2372      * socket means much for the paravirtualized PAPR platform) */
2373     return cpu_index / smp_threads / smp_cores;
2374 }
2375 
2376 static HotpluggableCPUList *spapr_query_hotpluggable_cpus(MachineState *machine)
2377 {
2378     int i;
2379     HotpluggableCPUList *head = NULL;
2380     sPAPRMachineState *spapr = SPAPR_MACHINE(machine);
2381     int spapr_max_cores = max_cpus / smp_threads;
2382 
2383     for (i = 0; i < spapr_max_cores; i++) {
2384         HotpluggableCPUList *list_item = g_new0(typeof(*list_item), 1);
2385         HotpluggableCPU *cpu_item = g_new0(typeof(*cpu_item), 1);
2386         CpuInstanceProperties *cpu_props = g_new0(typeof(*cpu_props), 1);
2387 
2388         cpu_item->type = spapr_get_cpu_core_type(machine->cpu_model);
2389         cpu_item->vcpus_count = smp_threads;
2390         cpu_props->has_core_id = true;
2391         cpu_props->core_id = i * smp_threads;
2392         /* TODO: add 'has_node/node' here to describe
2393            to which node core belongs */
2394 
2395         cpu_item->props = cpu_props;
2396         if (spapr->cores[i]) {
2397             cpu_item->has_qom_path = true;
2398             cpu_item->qom_path = object_get_canonical_path(spapr->cores[i]);
2399         }
2400         list_item->value = cpu_item;
2401         list_item->next = head;
2402         head = list_item;
2403     }
2404     return head;
2405 }
2406 
2407 static void spapr_machine_class_init(ObjectClass *oc, void *data)
2408 {
2409     MachineClass *mc = MACHINE_CLASS(oc);
2410     sPAPRMachineClass *smc = SPAPR_MACHINE_CLASS(oc);
2411     FWPathProviderClass *fwc = FW_PATH_PROVIDER_CLASS(oc);
2412     NMIClass *nc = NMI_CLASS(oc);
2413     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
2414 
2415     mc->desc = "pSeries Logical Partition (PAPR compliant)";
2416 
2417     /*
2418      * We set up the default / latest behaviour here.  The class_init
2419      * functions for the specific versioned machine types can override
2420      * these details for backwards compatibility
2421      */
2422     mc->init = ppc_spapr_init;
2423     mc->reset = ppc_spapr_reset;
2424     mc->block_default_type = IF_SCSI;
2425     mc->max_cpus = MAX_CPUMASK_BITS;
2426     mc->no_parallel = 1;
2427     mc->default_boot_order = "";
2428     mc->default_ram_size = 512 * M_BYTE;
2429     mc->kvm_type = spapr_kvm_type;
2430     mc->has_dynamic_sysbus = true;
2431     mc->pci_allow_0_address = true;
2432     mc->get_hotplug_handler = spapr_get_hotpug_handler;
2433     hc->pre_plug = spapr_machine_device_pre_plug;
2434     hc->plug = spapr_machine_device_plug;
2435     hc->unplug = spapr_machine_device_unplug;
2436     mc->cpu_index_to_socket_id = spapr_cpu_index_to_socket_id;
2437 
2438     smc->dr_lmb_enabled = true;
2439     mc->query_hotpluggable_cpus = spapr_query_hotpluggable_cpus;
2440     fwc->get_dev_path = spapr_get_fw_dev_path;
2441     nc->nmi_monitor_handler = spapr_nmi;
2442 }
2443 
2444 static const TypeInfo spapr_machine_info = {
2445     .name          = TYPE_SPAPR_MACHINE,
2446     .parent        = TYPE_MACHINE,
2447     .abstract      = true,
2448     .instance_size = sizeof(sPAPRMachineState),
2449     .instance_init = spapr_machine_initfn,
2450     .instance_finalize = spapr_machine_finalizefn,
2451     .class_size    = sizeof(sPAPRMachineClass),
2452     .class_init    = spapr_machine_class_init,
2453     .interfaces = (InterfaceInfo[]) {
2454         { TYPE_FW_PATH_PROVIDER },
2455         { TYPE_NMI },
2456         { TYPE_HOTPLUG_HANDLER },
2457         { }
2458     },
2459 };
2460 
2461 #define DEFINE_SPAPR_MACHINE(suffix, verstr, latest)                 \
2462     static void spapr_machine_##suffix##_class_init(ObjectClass *oc, \
2463                                                     void *data)      \
2464     {                                                                \
2465         MachineClass *mc = MACHINE_CLASS(oc);                        \
2466         spapr_machine_##suffix##_class_options(mc);                  \
2467         if (latest) {                                                \
2468             mc->alias = "pseries";                                   \
2469             mc->is_default = 1;                                      \
2470         }                                                            \
2471     }                                                                \
2472     static void spapr_machine_##suffix##_instance_init(Object *obj)  \
2473     {                                                                \
2474         MachineState *machine = MACHINE(obj);                        \
2475         spapr_machine_##suffix##_instance_options(machine);          \
2476     }                                                                \
2477     static const TypeInfo spapr_machine_##suffix##_info = {          \
2478         .name = MACHINE_TYPE_NAME("pseries-" verstr),                \
2479         .parent = TYPE_SPAPR_MACHINE,                                \
2480         .class_init = spapr_machine_##suffix##_class_init,           \
2481         .instance_init = spapr_machine_##suffix##_instance_init,     \
2482     };                                                               \
2483     static void spapr_machine_register_##suffix(void)                \
2484     {                                                                \
2485         type_register(&spapr_machine_##suffix##_info);               \
2486     }                                                                \
2487     type_init(spapr_machine_register_##suffix)
2488 
2489 /*
2490  * pseries-2.7
2491  */
2492 static void spapr_machine_2_7_instance_options(MachineState *machine)
2493 {
2494 }
2495 
2496 static void spapr_machine_2_7_class_options(MachineClass *mc)
2497 {
2498     /* Defaults for the latest behaviour inherited from the base class */
2499 }
2500 
2501 DEFINE_SPAPR_MACHINE(2_7, "2.7", true);
2502 
2503 /*
2504  * pseries-2.6
2505  */
2506 #define SPAPR_COMPAT_2_6 \
2507     HW_COMPAT_2_6 \
2508     { \
2509         .driver   = TYPE_SPAPR_PCI_HOST_BRIDGE,\
2510         .property = "ddw",\
2511         .value    = stringify(off),\
2512     },
2513 
2514 static void spapr_machine_2_6_instance_options(MachineState *machine)
2515 {
2516 }
2517 
2518 static void spapr_machine_2_6_class_options(MachineClass *mc)
2519 {
2520     spapr_machine_2_7_class_options(mc);
2521     mc->query_hotpluggable_cpus = NULL;
2522     SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_6);
2523 }
2524 
2525 DEFINE_SPAPR_MACHINE(2_6, "2.6", false);
2526 
2527 /*
2528  * pseries-2.5
2529  */
2530 #define SPAPR_COMPAT_2_5 \
2531     HW_COMPAT_2_5 \
2532     { \
2533         .driver   = "spapr-vlan", \
2534         .property = "use-rx-buffer-pools", \
2535         .value    = "off", \
2536     },
2537 
2538 static void spapr_machine_2_5_instance_options(MachineState *machine)
2539 {
2540 }
2541 
2542 static void spapr_machine_2_5_class_options(MachineClass *mc)
2543 {
2544     sPAPRMachineClass *smc = SPAPR_MACHINE_CLASS(mc);
2545 
2546     spapr_machine_2_6_class_options(mc);
2547     smc->use_ohci_by_default = true;
2548     SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_5);
2549 }
2550 
2551 DEFINE_SPAPR_MACHINE(2_5, "2.5", false);
2552 
2553 /*
2554  * pseries-2.4
2555  */
2556 #define SPAPR_COMPAT_2_4 \
2557         HW_COMPAT_2_4
2558 
2559 static void spapr_machine_2_4_instance_options(MachineState *machine)
2560 {
2561     spapr_machine_2_5_instance_options(machine);
2562 }
2563 
2564 static void spapr_machine_2_4_class_options(MachineClass *mc)
2565 {
2566     sPAPRMachineClass *smc = SPAPR_MACHINE_CLASS(mc);
2567 
2568     spapr_machine_2_5_class_options(mc);
2569     smc->dr_lmb_enabled = false;
2570     SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_4);
2571 }
2572 
2573 DEFINE_SPAPR_MACHINE(2_4, "2.4", false);
2574 
2575 /*
2576  * pseries-2.3
2577  */
2578 #define SPAPR_COMPAT_2_3 \
2579         HW_COMPAT_2_3 \
2580         {\
2581             .driver   = "spapr-pci-host-bridge",\
2582             .property = "dynamic-reconfiguration",\
2583             .value    = "off",\
2584         },
2585 
2586 static void spapr_machine_2_3_instance_options(MachineState *machine)
2587 {
2588     spapr_machine_2_4_instance_options(machine);
2589     savevm_skip_section_footers();
2590     global_state_set_optional();
2591     savevm_skip_configuration();
2592 }
2593 
2594 static void spapr_machine_2_3_class_options(MachineClass *mc)
2595 {
2596     spapr_machine_2_4_class_options(mc);
2597     SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_3);
2598 }
2599 DEFINE_SPAPR_MACHINE(2_3, "2.3", false);
2600 
2601 /*
2602  * pseries-2.2
2603  */
2604 
2605 #define SPAPR_COMPAT_2_2 \
2606         HW_COMPAT_2_2 \
2607         {\
2608             .driver   = TYPE_SPAPR_PCI_HOST_BRIDGE,\
2609             .property = "mem_win_size",\
2610             .value    = "0x20000000",\
2611         },
2612 
2613 static void spapr_machine_2_2_instance_options(MachineState *machine)
2614 {
2615     spapr_machine_2_3_instance_options(machine);
2616     machine->suppress_vmdesc = true;
2617 }
2618 
2619 static void spapr_machine_2_2_class_options(MachineClass *mc)
2620 {
2621     spapr_machine_2_3_class_options(mc);
2622     SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_2);
2623 }
2624 DEFINE_SPAPR_MACHINE(2_2, "2.2", false);
2625 
2626 /*
2627  * pseries-2.1
2628  */
2629 #define SPAPR_COMPAT_2_1 \
2630         HW_COMPAT_2_1
2631 
2632 static void spapr_machine_2_1_instance_options(MachineState *machine)
2633 {
2634     spapr_machine_2_2_instance_options(machine);
2635 }
2636 
2637 static void spapr_machine_2_1_class_options(MachineClass *mc)
2638 {
2639     spapr_machine_2_2_class_options(mc);
2640     SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_1);
2641 }
2642 DEFINE_SPAPR_MACHINE(2_1, "2.1", false);
2643 
2644 static void spapr_machine_register_types(void)
2645 {
2646     type_register_static(&spapr_machine_info);
2647 }
2648 
2649 type_init(spapr_machine_register_types)
2650