xref: /openbmc/qemu/hw/ppc/spapr.c (revision 10500ce2)
1 /*
2  * QEMU PowerPC pSeries Logical Partition (aka sPAPR) hardware System Emulator
3  *
4  * Copyright (c) 2004-2007 Fabrice Bellard
5  * Copyright (c) 2007 Jocelyn Mayer
6  * Copyright (c) 2010 David Gibson, IBM Corporation.
7  *
8  * Permission is hereby granted, free of charge, to any person obtaining a copy
9  * of this software and associated documentation files (the "Software"), to deal
10  * in the Software without restriction, including without limitation the rights
11  * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
12  * copies of the Software, and to permit persons to whom the Software is
13  * furnished to do so, subject to the following conditions:
14  *
15  * The above copyright notice and this permission notice shall be included in
16  * all copies or substantial portions of the Software.
17  *
18  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
19  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
20  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
21  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
22  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
23  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
24  * THE SOFTWARE.
25  *
26  */
27 #include "sysemu/sysemu.h"
28 #include "sysemu/numa.h"
29 #include "hw/hw.h"
30 #include "hw/fw-path-provider.h"
31 #include "elf.h"
32 #include "net/net.h"
33 #include "sysemu/block-backend.h"
34 #include "sysemu/cpus.h"
35 #include "sysemu/kvm.h"
36 #include "kvm_ppc.h"
37 #include "mmu-hash64.h"
38 #include "qom/cpu.h"
39 
40 #include "hw/boards.h"
41 #include "hw/ppc/ppc.h"
42 #include "hw/loader.h"
43 
44 #include "hw/ppc/spapr.h"
45 #include "hw/ppc/spapr_vio.h"
46 #include "hw/pci-host/spapr.h"
47 #include "hw/ppc/xics.h"
48 #include "hw/pci/msi.h"
49 
50 #include "hw/pci/pci.h"
51 #include "hw/scsi/scsi.h"
52 #include "hw/virtio/virtio-scsi.h"
53 
54 #include "exec/address-spaces.h"
55 #include "hw/usb.h"
56 #include "qemu/config-file.h"
57 #include "qemu/error-report.h"
58 #include "trace.h"
59 #include "hw/nmi.h"
60 
61 #include "hw/compat.h"
62 
63 #include <libfdt.h>
64 
65 /* SLOF memory layout:
66  *
67  * SLOF raw image loaded at 0, copies its romfs right below the flat
68  * device-tree, then position SLOF itself 31M below that
69  *
70  * So we set FW_OVERHEAD to 40MB which should account for all of that
71  * and more
72  *
73  * We load our kernel at 4M, leaving space for SLOF initial image
74  */
75 #define FDT_MAX_SIZE            0x40000
76 #define RTAS_MAX_SIZE           0x10000
77 #define RTAS_MAX_ADDR           0x80000000 /* RTAS must stay below that */
78 #define FW_MAX_SIZE             0x400000
79 #define FW_FILE_NAME            "slof.bin"
80 #define FW_OVERHEAD             0x2800000
81 #define KERNEL_LOAD_ADDR        FW_MAX_SIZE
82 
83 #define MIN_RMA_SLOF            128UL
84 
85 #define TIMEBASE_FREQ           512000000ULL
86 
87 #define MAX_CPUS                255
88 
89 #define PHANDLE_XICP            0x00001111
90 
91 #define HTAB_SIZE(spapr)        (1ULL << ((spapr)->htab_shift))
92 
93 typedef struct sPAPRMachineState sPAPRMachineState;
94 
95 #define TYPE_SPAPR_MACHINE      "spapr-machine"
96 #define SPAPR_MACHINE(obj) \
97     OBJECT_CHECK(sPAPRMachineState, (obj), TYPE_SPAPR_MACHINE)
98 
99 /**
100  * sPAPRMachineState:
101  */
102 struct sPAPRMachineState {
103     /*< private >*/
104     MachineState parent_obj;
105 
106     /*< public >*/
107     char *kvm_type;
108 };
109 
110 sPAPREnvironment *spapr;
111 
112 static XICSState *try_create_xics(const char *type, int nr_servers,
113                                   int nr_irqs, Error **errp)
114 {
115     Error *err = NULL;
116     DeviceState *dev;
117 
118     dev = qdev_create(NULL, type);
119     qdev_prop_set_uint32(dev, "nr_servers", nr_servers);
120     qdev_prop_set_uint32(dev, "nr_irqs", nr_irqs);
121     object_property_set_bool(OBJECT(dev), true, "realized", &err);
122     if (err) {
123         error_propagate(errp, err);
124         object_unparent(OBJECT(dev));
125         return NULL;
126     }
127     return XICS_COMMON(dev);
128 }
129 
130 static XICSState *xics_system_init(MachineState *machine,
131                                    int nr_servers, int nr_irqs)
132 {
133     XICSState *icp = NULL;
134 
135     if (kvm_enabled()) {
136         Error *err = NULL;
137 
138         if (machine_kernel_irqchip_allowed(machine)) {
139             icp = try_create_xics(TYPE_KVM_XICS, nr_servers, nr_irqs, &err);
140         }
141         if (machine_kernel_irqchip_required(machine) && !icp) {
142             error_report("kernel_irqchip requested but unavailable: %s",
143                          error_get_pretty(err));
144         }
145     }
146 
147     if (!icp) {
148         icp = try_create_xics(TYPE_XICS, nr_servers, nr_irqs, &error_abort);
149     }
150 
151     return icp;
152 }
153 
154 static int spapr_fixup_cpu_smt_dt(void *fdt, int offset, PowerPCCPU *cpu,
155                                   int smt_threads)
156 {
157     int i, ret = 0;
158     uint32_t servers_prop[smt_threads];
159     uint32_t gservers_prop[smt_threads * 2];
160     int index = ppc_get_vcpu_dt_id(cpu);
161 
162     if (cpu->cpu_version) {
163         ret = fdt_setprop_cell(fdt, offset, "cpu-version", cpu->cpu_version);
164         if (ret < 0) {
165             return ret;
166         }
167     }
168 
169     /* Build interrupt servers and gservers properties */
170     for (i = 0; i < smt_threads; i++) {
171         servers_prop[i] = cpu_to_be32(index + i);
172         /* Hack, direct the group queues back to cpu 0 */
173         gservers_prop[i*2] = cpu_to_be32(index + i);
174         gservers_prop[i*2 + 1] = 0;
175     }
176     ret = fdt_setprop(fdt, offset, "ibm,ppc-interrupt-server#s",
177                       servers_prop, sizeof(servers_prop));
178     if (ret < 0) {
179         return ret;
180     }
181     ret = fdt_setprop(fdt, offset, "ibm,ppc-interrupt-gserver#s",
182                       gservers_prop, sizeof(gservers_prop));
183 
184     return ret;
185 }
186 
187 static int spapr_fixup_cpu_dt(void *fdt, sPAPREnvironment *spapr)
188 {
189     int ret = 0, offset, cpus_offset;
190     CPUState *cs;
191     char cpu_model[32];
192     int smt = kvmppc_smt_threads();
193     uint32_t pft_size_prop[] = {0, cpu_to_be32(spapr->htab_shift)};
194 
195     CPU_FOREACH(cs) {
196         PowerPCCPU *cpu = POWERPC_CPU(cs);
197         DeviceClass *dc = DEVICE_GET_CLASS(cs);
198         int index = ppc_get_vcpu_dt_id(cpu);
199         uint32_t associativity[] = {cpu_to_be32(0x5),
200                                     cpu_to_be32(0x0),
201                                     cpu_to_be32(0x0),
202                                     cpu_to_be32(0x0),
203                                     cpu_to_be32(cs->numa_node),
204                                     cpu_to_be32(index)};
205 
206         if ((index % smt) != 0) {
207             continue;
208         }
209 
210         snprintf(cpu_model, 32, "%s@%x", dc->fw_name, index);
211 
212         cpus_offset = fdt_path_offset(fdt, "/cpus");
213         if (cpus_offset < 0) {
214             cpus_offset = fdt_add_subnode(fdt, fdt_path_offset(fdt, "/"),
215                                           "cpus");
216             if (cpus_offset < 0) {
217                 return cpus_offset;
218             }
219         }
220         offset = fdt_subnode_offset(fdt, cpus_offset, cpu_model);
221         if (offset < 0) {
222             offset = fdt_add_subnode(fdt, cpus_offset, cpu_model);
223             if (offset < 0) {
224                 return offset;
225             }
226         }
227 
228         if (nb_numa_nodes > 1) {
229             ret = fdt_setprop(fdt, offset, "ibm,associativity", associativity,
230                               sizeof(associativity));
231             if (ret < 0) {
232                 return ret;
233             }
234         }
235 
236         ret = fdt_setprop(fdt, offset, "ibm,pft-size",
237                           pft_size_prop, sizeof(pft_size_prop));
238         if (ret < 0) {
239             return ret;
240         }
241 
242         ret = spapr_fixup_cpu_smt_dt(fdt, offset, cpu,
243                                      ppc_get_compat_smt_threads(cpu));
244         if (ret < 0) {
245             return ret;
246         }
247     }
248     return ret;
249 }
250 
251 
252 static size_t create_page_sizes_prop(CPUPPCState *env, uint32_t *prop,
253                                      size_t maxsize)
254 {
255     size_t maxcells = maxsize / sizeof(uint32_t);
256     int i, j, count;
257     uint32_t *p = prop;
258 
259     for (i = 0; i < PPC_PAGE_SIZES_MAX_SZ; i++) {
260         struct ppc_one_seg_page_size *sps = &env->sps.sps[i];
261 
262         if (!sps->page_shift) {
263             break;
264         }
265         for (count = 0; count < PPC_PAGE_SIZES_MAX_SZ; count++) {
266             if (sps->enc[count].page_shift == 0) {
267                 break;
268             }
269         }
270         if ((p - prop) >= (maxcells - 3 - count * 2)) {
271             break;
272         }
273         *(p++) = cpu_to_be32(sps->page_shift);
274         *(p++) = cpu_to_be32(sps->slb_enc);
275         *(p++) = cpu_to_be32(count);
276         for (j = 0; j < count; j++) {
277             *(p++) = cpu_to_be32(sps->enc[j].page_shift);
278             *(p++) = cpu_to_be32(sps->enc[j].pte_enc);
279         }
280     }
281 
282     return (p - prop) * sizeof(uint32_t);
283 }
284 
285 static hwaddr spapr_node0_size(void)
286 {
287     if (nb_numa_nodes) {
288         int i;
289         for (i = 0; i < nb_numa_nodes; ++i) {
290             if (numa_info[i].node_mem) {
291                 return MIN(pow2floor(numa_info[i].node_mem), ram_size);
292             }
293         }
294     }
295     return ram_size;
296 }
297 
298 #define _FDT(exp) \
299     do { \
300         int ret = (exp);                                           \
301         if (ret < 0) {                                             \
302             fprintf(stderr, "qemu: error creating device tree: %s: %s\n", \
303                     #exp, fdt_strerror(ret));                      \
304             exit(1);                                               \
305         }                                                          \
306     } while (0)
307 
308 static void add_str(GString *s, const gchar *s1)
309 {
310     g_string_append_len(s, s1, strlen(s1) + 1);
311 }
312 
313 static void *spapr_create_fdt_skel(hwaddr initrd_base,
314                                    hwaddr initrd_size,
315                                    hwaddr kernel_size,
316                                    bool little_endian,
317                                    const char *kernel_cmdline,
318                                    uint32_t epow_irq)
319 {
320     void *fdt;
321     CPUState *cs;
322     uint32_t start_prop = cpu_to_be32(initrd_base);
323     uint32_t end_prop = cpu_to_be32(initrd_base + initrd_size);
324     GString *hypertas = g_string_sized_new(256);
325     GString *qemu_hypertas = g_string_sized_new(256);
326     uint32_t refpoints[] = {cpu_to_be32(0x4), cpu_to_be32(0x4)};
327     uint32_t interrupt_server_ranges_prop[] = {0, cpu_to_be32(smp_cpus)};
328     int smt = kvmppc_smt_threads();
329     unsigned char vec5[] = {0x0, 0x0, 0x0, 0x0, 0x0, 0x80};
330     QemuOpts *opts = qemu_opts_find(qemu_find_opts("smp-opts"), NULL);
331     unsigned sockets = opts ? qemu_opt_get_number(opts, "sockets", 0) : 0;
332     uint32_t cpus_per_socket = sockets ? (smp_cpus / sockets) : 1;
333     char *buf;
334 
335     add_str(hypertas, "hcall-pft");
336     add_str(hypertas, "hcall-term");
337     add_str(hypertas, "hcall-dabr");
338     add_str(hypertas, "hcall-interrupt");
339     add_str(hypertas, "hcall-tce");
340     add_str(hypertas, "hcall-vio");
341     add_str(hypertas, "hcall-splpar");
342     add_str(hypertas, "hcall-bulk");
343     add_str(hypertas, "hcall-set-mode");
344     add_str(qemu_hypertas, "hcall-memop1");
345 
346     fdt = g_malloc0(FDT_MAX_SIZE);
347     _FDT((fdt_create(fdt, FDT_MAX_SIZE)));
348 
349     if (kernel_size) {
350         _FDT((fdt_add_reservemap_entry(fdt, KERNEL_LOAD_ADDR, kernel_size)));
351     }
352     if (initrd_size) {
353         _FDT((fdt_add_reservemap_entry(fdt, initrd_base, initrd_size)));
354     }
355     _FDT((fdt_finish_reservemap(fdt)));
356 
357     /* Root node */
358     _FDT((fdt_begin_node(fdt, "")));
359     _FDT((fdt_property_string(fdt, "device_type", "chrp")));
360     _FDT((fdt_property_string(fdt, "model", "IBM pSeries (emulated by qemu)")));
361     _FDT((fdt_property_string(fdt, "compatible", "qemu,pseries")));
362 
363     /*
364      * Add info to guest to indentify which host is it being run on
365      * and what is the uuid of the guest
366      */
367     if (kvmppc_get_host_model(&buf)) {
368         _FDT((fdt_property_string(fdt, "host-model", buf)));
369         g_free(buf);
370     }
371     if (kvmppc_get_host_serial(&buf)) {
372         _FDT((fdt_property_string(fdt, "host-serial", buf)));
373         g_free(buf);
374     }
375 
376     buf = g_strdup_printf(UUID_FMT, qemu_uuid[0], qemu_uuid[1],
377                           qemu_uuid[2], qemu_uuid[3], qemu_uuid[4],
378                           qemu_uuid[5], qemu_uuid[6], qemu_uuid[7],
379                           qemu_uuid[8], qemu_uuid[9], qemu_uuid[10],
380                           qemu_uuid[11], qemu_uuid[12], qemu_uuid[13],
381                           qemu_uuid[14], qemu_uuid[15]);
382 
383     _FDT((fdt_property_string(fdt, "vm,uuid", buf)));
384     g_free(buf);
385 
386     _FDT((fdt_property_cell(fdt, "#address-cells", 0x2)));
387     _FDT((fdt_property_cell(fdt, "#size-cells", 0x2)));
388 
389     /* /chosen */
390     _FDT((fdt_begin_node(fdt, "chosen")));
391 
392     /* Set Form1_affinity */
393     _FDT((fdt_property(fdt, "ibm,architecture-vec-5", vec5, sizeof(vec5))));
394 
395     _FDT((fdt_property_string(fdt, "bootargs", kernel_cmdline)));
396     _FDT((fdt_property(fdt, "linux,initrd-start",
397                        &start_prop, sizeof(start_prop))));
398     _FDT((fdt_property(fdt, "linux,initrd-end",
399                        &end_prop, sizeof(end_prop))));
400     if (kernel_size) {
401         uint64_t kprop[2] = { cpu_to_be64(KERNEL_LOAD_ADDR),
402                               cpu_to_be64(kernel_size) };
403 
404         _FDT((fdt_property(fdt, "qemu,boot-kernel", &kprop, sizeof(kprop))));
405         if (little_endian) {
406             _FDT((fdt_property(fdt, "qemu,boot-kernel-le", NULL, 0)));
407         }
408     }
409     if (boot_menu) {
410         _FDT((fdt_property_cell(fdt, "qemu,boot-menu", boot_menu)));
411     }
412     _FDT((fdt_property_cell(fdt, "qemu,graphic-width", graphic_width)));
413     _FDT((fdt_property_cell(fdt, "qemu,graphic-height", graphic_height)));
414     _FDT((fdt_property_cell(fdt, "qemu,graphic-depth", graphic_depth)));
415 
416     _FDT((fdt_end_node(fdt)));
417 
418     /* cpus */
419     _FDT((fdt_begin_node(fdt, "cpus")));
420 
421     _FDT((fdt_property_cell(fdt, "#address-cells", 0x1)));
422     _FDT((fdt_property_cell(fdt, "#size-cells", 0x0)));
423 
424     CPU_FOREACH(cs) {
425         PowerPCCPU *cpu = POWERPC_CPU(cs);
426         CPUPPCState *env = &cpu->env;
427         DeviceClass *dc = DEVICE_GET_CLASS(cs);
428         PowerPCCPUClass *pcc = POWERPC_CPU_GET_CLASS(cs);
429         int index = ppc_get_vcpu_dt_id(cpu);
430         char *nodename;
431         uint32_t segs[] = {cpu_to_be32(28), cpu_to_be32(40),
432                            0xffffffff, 0xffffffff};
433         uint32_t tbfreq = kvm_enabled() ? kvmppc_get_tbfreq() : TIMEBASE_FREQ;
434         uint32_t cpufreq = kvm_enabled() ? kvmppc_get_clockfreq() : 1000000000;
435         uint32_t page_sizes_prop[64];
436         size_t page_sizes_prop_size;
437 
438         if ((index % smt) != 0) {
439             continue;
440         }
441 
442         nodename = g_strdup_printf("%s@%x", dc->fw_name, index);
443 
444         _FDT((fdt_begin_node(fdt, nodename)));
445 
446         g_free(nodename);
447 
448         _FDT((fdt_property_cell(fdt, "reg", index)));
449         _FDT((fdt_property_string(fdt, "device_type", "cpu")));
450 
451         _FDT((fdt_property_cell(fdt, "cpu-version", env->spr[SPR_PVR])));
452         _FDT((fdt_property_cell(fdt, "d-cache-block-size",
453                                 env->dcache_line_size)));
454         _FDT((fdt_property_cell(fdt, "d-cache-line-size",
455                                 env->dcache_line_size)));
456         _FDT((fdt_property_cell(fdt, "i-cache-block-size",
457                                 env->icache_line_size)));
458         _FDT((fdt_property_cell(fdt, "i-cache-line-size",
459                                 env->icache_line_size)));
460 
461         if (pcc->l1_dcache_size) {
462             _FDT((fdt_property_cell(fdt, "d-cache-size", pcc->l1_dcache_size)));
463         } else {
464             fprintf(stderr, "Warning: Unknown L1 dcache size for cpu\n");
465         }
466         if (pcc->l1_icache_size) {
467             _FDT((fdt_property_cell(fdt, "i-cache-size", pcc->l1_icache_size)));
468         } else {
469             fprintf(stderr, "Warning: Unknown L1 icache size for cpu\n");
470         }
471 
472         _FDT((fdt_property_cell(fdt, "timebase-frequency", tbfreq)));
473         _FDT((fdt_property_cell(fdt, "clock-frequency", cpufreq)));
474         _FDT((fdt_property_cell(fdt, "ibm,slb-size", env->slb_nr)));
475         _FDT((fdt_property_string(fdt, "status", "okay")));
476         _FDT((fdt_property(fdt, "64-bit", NULL, 0)));
477 
478         if (env->spr_cb[SPR_PURR].oea_read) {
479             _FDT((fdt_property(fdt, "ibm,purr", NULL, 0)));
480         }
481 
482         if (env->mmu_model & POWERPC_MMU_1TSEG) {
483             _FDT((fdt_property(fdt, "ibm,processor-segment-sizes",
484                                segs, sizeof(segs))));
485         }
486 
487         /* Advertise VMX/VSX (vector extensions) if available
488          *   0 / no property == no vector extensions
489          *   1               == VMX / Altivec available
490          *   2               == VSX available */
491         if (env->insns_flags & PPC_ALTIVEC) {
492             uint32_t vmx = (env->insns_flags2 & PPC2_VSX) ? 2 : 1;
493 
494             _FDT((fdt_property_cell(fdt, "ibm,vmx", vmx)));
495         }
496 
497         /* Advertise DFP (Decimal Floating Point) if available
498          *   0 / no property == no DFP
499          *   1               == DFP available */
500         if (env->insns_flags2 & PPC2_DFP) {
501             _FDT((fdt_property_cell(fdt, "ibm,dfp", 1)));
502         }
503 
504         page_sizes_prop_size = create_page_sizes_prop(env, page_sizes_prop,
505                                                       sizeof(page_sizes_prop));
506         if (page_sizes_prop_size) {
507             _FDT((fdt_property(fdt, "ibm,segment-page-sizes",
508                                page_sizes_prop, page_sizes_prop_size)));
509         }
510 
511         _FDT((fdt_property_cell(fdt, "ibm,chip-id",
512                                 cs->cpu_index / cpus_per_socket)));
513 
514         _FDT((fdt_end_node(fdt)));
515     }
516 
517     _FDT((fdt_end_node(fdt)));
518 
519     /* RTAS */
520     _FDT((fdt_begin_node(fdt, "rtas")));
521 
522     if (!kvm_enabled() || kvmppc_spapr_use_multitce()) {
523         add_str(hypertas, "hcall-multi-tce");
524     }
525     _FDT((fdt_property(fdt, "ibm,hypertas-functions", hypertas->str,
526                        hypertas->len)));
527     g_string_free(hypertas, TRUE);
528     _FDT((fdt_property(fdt, "qemu,hypertas-functions", qemu_hypertas->str,
529                        qemu_hypertas->len)));
530     g_string_free(qemu_hypertas, TRUE);
531 
532     _FDT((fdt_property(fdt, "ibm,associativity-reference-points",
533         refpoints, sizeof(refpoints))));
534 
535     _FDT((fdt_property_cell(fdt, "rtas-error-log-max", RTAS_ERROR_LOG_MAX)));
536     _FDT((fdt_property_cell(fdt, "rtas-event-scan-rate",
537                             RTAS_EVENT_SCAN_RATE)));
538 
539     /*
540      * According to PAPR, rtas ibm,os-term does not guarantee a return
541      * back to the guest cpu.
542      *
543      * While an additional ibm,extended-os-term property indicates that
544      * rtas call return will always occur. Set this property.
545      */
546     _FDT((fdt_property(fdt, "ibm,extended-os-term", NULL, 0)));
547 
548     _FDT((fdt_end_node(fdt)));
549 
550     /* interrupt controller */
551     _FDT((fdt_begin_node(fdt, "interrupt-controller")));
552 
553     _FDT((fdt_property_string(fdt, "device_type",
554                               "PowerPC-External-Interrupt-Presentation")));
555     _FDT((fdt_property_string(fdt, "compatible", "IBM,ppc-xicp")));
556     _FDT((fdt_property(fdt, "interrupt-controller", NULL, 0)));
557     _FDT((fdt_property(fdt, "ibm,interrupt-server-ranges",
558                        interrupt_server_ranges_prop,
559                        sizeof(interrupt_server_ranges_prop))));
560     _FDT((fdt_property_cell(fdt, "#interrupt-cells", 2)));
561     _FDT((fdt_property_cell(fdt, "linux,phandle", PHANDLE_XICP)));
562     _FDT((fdt_property_cell(fdt, "phandle", PHANDLE_XICP)));
563 
564     _FDT((fdt_end_node(fdt)));
565 
566     /* vdevice */
567     _FDT((fdt_begin_node(fdt, "vdevice")));
568 
569     _FDT((fdt_property_string(fdt, "device_type", "vdevice")));
570     _FDT((fdt_property_string(fdt, "compatible", "IBM,vdevice")));
571     _FDT((fdt_property_cell(fdt, "#address-cells", 0x1)));
572     _FDT((fdt_property_cell(fdt, "#size-cells", 0x0)));
573     _FDT((fdt_property_cell(fdt, "#interrupt-cells", 0x2)));
574     _FDT((fdt_property(fdt, "interrupt-controller", NULL, 0)));
575 
576     _FDT((fdt_end_node(fdt)));
577 
578     /* event-sources */
579     spapr_events_fdt_skel(fdt, epow_irq);
580 
581     /* /hypervisor node */
582     if (kvm_enabled()) {
583         uint8_t hypercall[16];
584 
585         /* indicate KVM hypercall interface */
586         _FDT((fdt_begin_node(fdt, "hypervisor")));
587         _FDT((fdt_property_string(fdt, "compatible", "linux,kvm")));
588         if (kvmppc_has_cap_fixup_hcalls()) {
589             /*
590              * Older KVM versions with older guest kernels were broken with the
591              * magic page, don't allow the guest to map it.
592              */
593             kvmppc_get_hypercall(first_cpu->env_ptr, hypercall,
594                                  sizeof(hypercall));
595             _FDT((fdt_property(fdt, "hcall-instructions", hypercall,
596                               sizeof(hypercall))));
597         }
598         _FDT((fdt_end_node(fdt)));
599     }
600 
601     _FDT((fdt_end_node(fdt))); /* close root node */
602     _FDT((fdt_finish(fdt)));
603 
604     return fdt;
605 }
606 
607 int spapr_h_cas_compose_response(target_ulong addr, target_ulong size)
608 {
609     void *fdt, *fdt_skel;
610     sPAPRDeviceTreeUpdateHeader hdr = { .version_id = 1 };
611 
612     size -= sizeof(hdr);
613 
614     /* Create sceleton */
615     fdt_skel = g_malloc0(size);
616     _FDT((fdt_create(fdt_skel, size)));
617     _FDT((fdt_begin_node(fdt_skel, "")));
618     _FDT((fdt_end_node(fdt_skel)));
619     _FDT((fdt_finish(fdt_skel)));
620     fdt = g_malloc0(size);
621     _FDT((fdt_open_into(fdt_skel, fdt, size)));
622     g_free(fdt_skel);
623 
624     /* Fix skeleton up */
625     _FDT((spapr_fixup_cpu_dt(fdt, spapr)));
626 
627     /* Pack resulting tree */
628     _FDT((fdt_pack(fdt)));
629 
630     if (fdt_totalsize(fdt) + sizeof(hdr) > size) {
631         trace_spapr_cas_failed(size);
632         return -1;
633     }
634 
635     cpu_physical_memory_write(addr, &hdr, sizeof(hdr));
636     cpu_physical_memory_write(addr + sizeof(hdr), fdt, fdt_totalsize(fdt));
637     trace_spapr_cas_continue(fdt_totalsize(fdt) + sizeof(hdr));
638     g_free(fdt);
639 
640     return 0;
641 }
642 
643 static void spapr_populate_memory_node(void *fdt, int nodeid, hwaddr start,
644                                        hwaddr size)
645 {
646     uint32_t associativity[] = {
647         cpu_to_be32(0x4), /* length */
648         cpu_to_be32(0x0), cpu_to_be32(0x0),
649         cpu_to_be32(0x0), cpu_to_be32(nodeid)
650     };
651     char mem_name[32];
652     uint64_t mem_reg_property[2];
653     int off;
654 
655     mem_reg_property[0] = cpu_to_be64(start);
656     mem_reg_property[1] = cpu_to_be64(size);
657 
658     sprintf(mem_name, "memory@" TARGET_FMT_lx, start);
659     off = fdt_add_subnode(fdt, 0, mem_name);
660     _FDT(off);
661     _FDT((fdt_setprop_string(fdt, off, "device_type", "memory")));
662     _FDT((fdt_setprop(fdt, off, "reg", mem_reg_property,
663                       sizeof(mem_reg_property))));
664     _FDT((fdt_setprop(fdt, off, "ibm,associativity", associativity,
665                       sizeof(associativity))));
666 }
667 
668 static int spapr_populate_memory(sPAPREnvironment *spapr, void *fdt)
669 {
670     hwaddr mem_start, node_size;
671     int i, nb_nodes = nb_numa_nodes;
672     NodeInfo *nodes = numa_info;
673     NodeInfo ramnode;
674 
675     /* No NUMA nodes, assume there is just one node with whole RAM */
676     if (!nb_numa_nodes) {
677         nb_nodes = 1;
678         ramnode.node_mem = ram_size;
679         nodes = &ramnode;
680     }
681 
682     for (i = 0, mem_start = 0; i < nb_nodes; ++i) {
683         if (!nodes[i].node_mem) {
684             continue;
685         }
686         if (mem_start >= ram_size) {
687             node_size = 0;
688         } else {
689             node_size = nodes[i].node_mem;
690             if (node_size > ram_size - mem_start) {
691                 node_size = ram_size - mem_start;
692             }
693         }
694         if (!mem_start) {
695             /* ppc_spapr_init() checks for rma_size <= node0_size already */
696             spapr_populate_memory_node(fdt, i, 0, spapr->rma_size);
697             mem_start += spapr->rma_size;
698             node_size -= spapr->rma_size;
699         }
700         for ( ; node_size; ) {
701             hwaddr sizetmp = pow2floor(node_size);
702 
703             /* mem_start != 0 here */
704             if (ctzl(mem_start) < ctzl(sizetmp)) {
705                 sizetmp = 1ULL << ctzl(mem_start);
706             }
707 
708             spapr_populate_memory_node(fdt, i, mem_start, sizetmp);
709             node_size -= sizetmp;
710             mem_start += sizetmp;
711         }
712     }
713 
714     return 0;
715 }
716 
717 static void spapr_finalize_fdt(sPAPREnvironment *spapr,
718                                hwaddr fdt_addr,
719                                hwaddr rtas_addr,
720                                hwaddr rtas_size)
721 {
722     MachineState *machine = MACHINE(qdev_get_machine());
723     const char *boot_device = machine->boot_order;
724     int ret, i;
725     size_t cb = 0;
726     char *bootlist;
727     void *fdt;
728     sPAPRPHBState *phb;
729 
730     fdt = g_malloc(FDT_MAX_SIZE);
731 
732     /* open out the base tree into a temp buffer for the final tweaks */
733     _FDT((fdt_open_into(spapr->fdt_skel, fdt, FDT_MAX_SIZE)));
734 
735     ret = spapr_populate_memory(spapr, fdt);
736     if (ret < 0) {
737         fprintf(stderr, "couldn't setup memory nodes in fdt\n");
738         exit(1);
739     }
740 
741     ret = spapr_populate_vdevice(spapr->vio_bus, fdt);
742     if (ret < 0) {
743         fprintf(stderr, "couldn't setup vio devices in fdt\n");
744         exit(1);
745     }
746 
747     QLIST_FOREACH(phb, &spapr->phbs, list) {
748         ret = spapr_populate_pci_dt(phb, PHANDLE_XICP, fdt);
749     }
750 
751     if (ret < 0) {
752         fprintf(stderr, "couldn't setup PCI devices in fdt\n");
753         exit(1);
754     }
755 
756     /* RTAS */
757     ret = spapr_rtas_device_tree_setup(fdt, rtas_addr, rtas_size);
758     if (ret < 0) {
759         fprintf(stderr, "Couldn't set up RTAS device tree properties\n");
760     }
761 
762     /* Advertise NUMA via ibm,associativity */
763     ret = spapr_fixup_cpu_dt(fdt, spapr);
764     if (ret < 0) {
765         fprintf(stderr, "Couldn't finalize CPU device tree properties\n");
766     }
767 
768     bootlist = get_boot_devices_list(&cb, true);
769     if (cb && bootlist) {
770         int offset = fdt_path_offset(fdt, "/chosen");
771         if (offset < 0) {
772             exit(1);
773         }
774         for (i = 0; i < cb; i++) {
775             if (bootlist[i] == '\n') {
776                 bootlist[i] = ' ';
777             }
778 
779         }
780         ret = fdt_setprop_string(fdt, offset, "qemu,boot-list", bootlist);
781     }
782 
783     if (boot_device && strlen(boot_device)) {
784         int offset = fdt_path_offset(fdt, "/chosen");
785 
786         if (offset < 0) {
787             exit(1);
788         }
789         fdt_setprop_string(fdt, offset, "qemu,boot-device", boot_device);
790     }
791 
792     if (!spapr->has_graphics) {
793         spapr_populate_chosen_stdout(fdt, spapr->vio_bus);
794     }
795 
796     _FDT((fdt_pack(fdt)));
797 
798     if (fdt_totalsize(fdt) > FDT_MAX_SIZE) {
799         error_report("FDT too big ! 0x%x bytes (max is 0x%x)",
800                      fdt_totalsize(fdt), FDT_MAX_SIZE);
801         exit(1);
802     }
803 
804     cpu_physical_memory_write(fdt_addr, fdt, fdt_totalsize(fdt));
805 
806     g_free(bootlist);
807     g_free(fdt);
808 }
809 
810 static uint64_t translate_kernel_address(void *opaque, uint64_t addr)
811 {
812     return (addr & 0x0fffffff) + KERNEL_LOAD_ADDR;
813 }
814 
815 static void emulate_spapr_hypercall(PowerPCCPU *cpu)
816 {
817     CPUPPCState *env = &cpu->env;
818 
819     if (msr_pr) {
820         hcall_dprintf("Hypercall made with MSR[PR]=1\n");
821         env->gpr[3] = H_PRIVILEGE;
822     } else {
823         env->gpr[3] = spapr_hypercall(cpu, env->gpr[3], &env->gpr[4]);
824     }
825 }
826 
827 #define HPTE(_table, _i)   (void *)(((uint64_t *)(_table)) + ((_i) * 2))
828 #define HPTE_VALID(_hpte)  (tswap64(*((uint64_t *)(_hpte))) & HPTE64_V_VALID)
829 #define HPTE_DIRTY(_hpte)  (tswap64(*((uint64_t *)(_hpte))) & HPTE64_V_HPTE_DIRTY)
830 #define CLEAN_HPTE(_hpte)  ((*(uint64_t *)(_hpte)) &= tswap64(~HPTE64_V_HPTE_DIRTY))
831 #define DIRTY_HPTE(_hpte)  ((*(uint64_t *)(_hpte)) |= tswap64(HPTE64_V_HPTE_DIRTY))
832 
833 static void spapr_reset_htab(sPAPREnvironment *spapr)
834 {
835     long shift;
836     int index;
837 
838     /* allocate hash page table.  For now we always make this 16mb,
839      * later we should probably make it scale to the size of guest
840      * RAM */
841 
842     shift = kvmppc_reset_htab(spapr->htab_shift);
843 
844     if (shift > 0) {
845         /* Kernel handles htab, we don't need to allocate one */
846         spapr->htab_shift = shift;
847         kvmppc_kern_htab = true;
848 
849         /* Tell readers to update their file descriptor */
850         if (spapr->htab_fd >= 0) {
851             spapr->htab_fd_stale = true;
852         }
853     } else {
854         if (!spapr->htab) {
855             /* Allocate an htab if we don't yet have one */
856             spapr->htab = qemu_memalign(HTAB_SIZE(spapr), HTAB_SIZE(spapr));
857         }
858 
859         /* And clear it */
860         memset(spapr->htab, 0, HTAB_SIZE(spapr));
861 
862         for (index = 0; index < HTAB_SIZE(spapr) / HASH_PTE_SIZE_64; index++) {
863             DIRTY_HPTE(HPTE(spapr->htab, index));
864         }
865     }
866 
867     /* Update the RMA size if necessary */
868     if (spapr->vrma_adjust) {
869         spapr->rma_size = kvmppc_rma_size(spapr_node0_size(),
870                                           spapr->htab_shift);
871     }
872 }
873 
874 static int find_unknown_sysbus_device(SysBusDevice *sbdev, void *opaque)
875 {
876     bool matched = false;
877 
878     if (object_dynamic_cast(OBJECT(sbdev), TYPE_SPAPR_PCI_HOST_BRIDGE)) {
879         matched = true;
880     }
881 
882     if (!matched) {
883         error_report("Device %s is not supported by this machine yet.",
884                      qdev_fw_name(DEVICE(sbdev)));
885         exit(1);
886     }
887 
888     return 0;
889 }
890 
891 /*
892  * A guest reset will cause spapr->htab_fd to become stale if being used.
893  * Reopen the file descriptor to make sure the whole HTAB is properly read.
894  */
895 static int spapr_check_htab_fd(sPAPREnvironment *spapr)
896 {
897     int rc = 0;
898 
899     if (spapr->htab_fd_stale) {
900         close(spapr->htab_fd);
901         spapr->htab_fd = kvmppc_get_htab_fd(false);
902         if (spapr->htab_fd < 0) {
903             error_report("Unable to open fd for reading hash table from KVM: "
904                          "%s", strerror(errno));
905             rc = -1;
906         }
907         spapr->htab_fd_stale = false;
908     }
909 
910     return rc;
911 }
912 
913 static void ppc_spapr_reset(void)
914 {
915     PowerPCCPU *first_ppc_cpu;
916     uint32_t rtas_limit;
917 
918     /* Check for unknown sysbus devices */
919     foreach_dynamic_sysbus_device(find_unknown_sysbus_device, NULL);
920 
921     /* Reset the hash table & recalc the RMA */
922     spapr_reset_htab(spapr);
923 
924     qemu_devices_reset();
925 
926     /*
927      * We place the device tree and RTAS just below either the top of the RMA,
928      * or just below 2GB, whichever is lowere, so that it can be
929      * processed with 32-bit real mode code if necessary
930      */
931     rtas_limit = MIN(spapr->rma_size, RTAS_MAX_ADDR);
932     spapr->rtas_addr = rtas_limit - RTAS_MAX_SIZE;
933     spapr->fdt_addr = spapr->rtas_addr - FDT_MAX_SIZE;
934 
935     /* Load the fdt */
936     spapr_finalize_fdt(spapr, spapr->fdt_addr, spapr->rtas_addr,
937                        spapr->rtas_size);
938 
939     /* Copy RTAS over */
940     cpu_physical_memory_write(spapr->rtas_addr, spapr->rtas_blob,
941                               spapr->rtas_size);
942 
943     /* Set up the entry state */
944     first_ppc_cpu = POWERPC_CPU(first_cpu);
945     first_ppc_cpu->env.gpr[3] = spapr->fdt_addr;
946     first_ppc_cpu->env.gpr[5] = 0;
947     first_cpu->halted = 0;
948     first_ppc_cpu->env.nip = spapr->entry_point;
949 
950 }
951 
952 static void spapr_cpu_reset(void *opaque)
953 {
954     PowerPCCPU *cpu = opaque;
955     CPUState *cs = CPU(cpu);
956     CPUPPCState *env = &cpu->env;
957 
958     cpu_reset(cs);
959 
960     /* All CPUs start halted.  CPU0 is unhalted from the machine level
961      * reset code and the rest are explicitly started up by the guest
962      * using an RTAS call */
963     cs->halted = 1;
964 
965     env->spr[SPR_HIOR] = 0;
966 
967     env->external_htab = (uint8_t *)spapr->htab;
968     if (kvm_enabled() && !env->external_htab) {
969         /*
970          * HV KVM, set external_htab to 1 so our ppc_hash64_load_hpte*
971          * functions do the right thing.
972          */
973         env->external_htab = (void *)1;
974     }
975     env->htab_base = -1;
976     /*
977      * htab_mask is the mask used to normalize hash value to PTEG index.
978      * htab_shift is log2 of hash table size.
979      * We have 8 hpte per group, and each hpte is 16 bytes.
980      * ie have 128 bytes per hpte entry.
981      */
982     env->htab_mask = (1ULL << ((spapr)->htab_shift - 7)) - 1;
983     env->spr[SPR_SDR1] = (target_ulong)(uintptr_t)spapr->htab |
984         (spapr->htab_shift - 18);
985 }
986 
987 static void spapr_create_nvram(sPAPREnvironment *spapr)
988 {
989     DeviceState *dev = qdev_create(&spapr->vio_bus->bus, "spapr-nvram");
990     DriveInfo *dinfo = drive_get(IF_PFLASH, 0, 0);
991 
992     if (dinfo) {
993         qdev_prop_set_drive_nofail(dev, "drive", blk_by_legacy_dinfo(dinfo));
994     }
995 
996     qdev_init_nofail(dev);
997 
998     spapr->nvram = (struct sPAPRNVRAM *)dev;
999 }
1000 
1001 static void spapr_rtc_create(sPAPREnvironment *spapr)
1002 {
1003     DeviceState *dev = qdev_create(NULL, TYPE_SPAPR_RTC);
1004 
1005     qdev_init_nofail(dev);
1006     spapr->rtc = dev;
1007 
1008     object_property_add_alias(qdev_get_machine(), "rtc-time",
1009                               OBJECT(spapr->rtc), "date", NULL);
1010 }
1011 
1012 /* Returns whether we want to use VGA or not */
1013 static int spapr_vga_init(PCIBus *pci_bus)
1014 {
1015     switch (vga_interface_type) {
1016     case VGA_NONE:
1017         return false;
1018     case VGA_DEVICE:
1019         return true;
1020     case VGA_STD:
1021         return pci_vga_init(pci_bus) != NULL;
1022     default:
1023         fprintf(stderr, "This vga model is not supported,"
1024                 "currently it only supports -vga std\n");
1025         exit(0);
1026     }
1027 }
1028 
1029 static int spapr_post_load(void *opaque, int version_id)
1030 {
1031     sPAPREnvironment *spapr = (sPAPREnvironment *)opaque;
1032     int err = 0;
1033 
1034     /* In earlier versions, there was no separate qdev for the PAPR
1035      * RTC, so the RTC offset was stored directly in sPAPREnvironment.
1036      * So when migrating from those versions, poke the incoming offset
1037      * value into the RTC device */
1038     if (version_id < 3) {
1039         err = spapr_rtc_import_offset(spapr->rtc, spapr->rtc_offset);
1040     }
1041 
1042     return err;
1043 }
1044 
1045 static bool version_before_3(void *opaque, int version_id)
1046 {
1047     return version_id < 3;
1048 }
1049 
1050 static const VMStateDescription vmstate_spapr = {
1051     .name = "spapr",
1052     .version_id = 3,
1053     .minimum_version_id = 1,
1054     .post_load = spapr_post_load,
1055     .fields = (VMStateField[]) {
1056         /* used to be @next_irq */
1057         VMSTATE_UNUSED_BUFFER(version_before_3, 0, 4),
1058 
1059         /* RTC offset */
1060         VMSTATE_UINT64_TEST(rtc_offset, sPAPREnvironment, version_before_3),
1061 
1062         VMSTATE_PPC_TIMEBASE_V(tb, sPAPREnvironment, 2),
1063         VMSTATE_END_OF_LIST()
1064     },
1065 };
1066 
1067 static int htab_save_setup(QEMUFile *f, void *opaque)
1068 {
1069     sPAPREnvironment *spapr = opaque;
1070 
1071     /* "Iteration" header */
1072     qemu_put_be32(f, spapr->htab_shift);
1073 
1074     if (spapr->htab) {
1075         spapr->htab_save_index = 0;
1076         spapr->htab_first_pass = true;
1077     } else {
1078         assert(kvm_enabled());
1079 
1080         spapr->htab_fd = kvmppc_get_htab_fd(false);
1081         spapr->htab_fd_stale = false;
1082         if (spapr->htab_fd < 0) {
1083             fprintf(stderr, "Unable to open fd for reading hash table from KVM: %s\n",
1084                     strerror(errno));
1085             return -1;
1086         }
1087     }
1088 
1089 
1090     return 0;
1091 }
1092 
1093 static void htab_save_first_pass(QEMUFile *f, sPAPREnvironment *spapr,
1094                                  int64_t max_ns)
1095 {
1096     int htabslots = HTAB_SIZE(spapr) / HASH_PTE_SIZE_64;
1097     int index = spapr->htab_save_index;
1098     int64_t starttime = qemu_clock_get_ns(QEMU_CLOCK_REALTIME);
1099 
1100     assert(spapr->htab_first_pass);
1101 
1102     do {
1103         int chunkstart;
1104 
1105         /* Consume invalid HPTEs */
1106         while ((index < htabslots)
1107                && !HPTE_VALID(HPTE(spapr->htab, index))) {
1108             index++;
1109             CLEAN_HPTE(HPTE(spapr->htab, index));
1110         }
1111 
1112         /* Consume valid HPTEs */
1113         chunkstart = index;
1114         while ((index < htabslots) && (index - chunkstart < USHRT_MAX)
1115                && HPTE_VALID(HPTE(spapr->htab, index))) {
1116             index++;
1117             CLEAN_HPTE(HPTE(spapr->htab, index));
1118         }
1119 
1120         if (index > chunkstart) {
1121             int n_valid = index - chunkstart;
1122 
1123             qemu_put_be32(f, chunkstart);
1124             qemu_put_be16(f, n_valid);
1125             qemu_put_be16(f, 0);
1126             qemu_put_buffer(f, HPTE(spapr->htab, chunkstart),
1127                             HASH_PTE_SIZE_64 * n_valid);
1128 
1129             if ((qemu_clock_get_ns(QEMU_CLOCK_REALTIME) - starttime) > max_ns) {
1130                 break;
1131             }
1132         }
1133     } while ((index < htabslots) && !qemu_file_rate_limit(f));
1134 
1135     if (index >= htabslots) {
1136         assert(index == htabslots);
1137         index = 0;
1138         spapr->htab_first_pass = false;
1139     }
1140     spapr->htab_save_index = index;
1141 }
1142 
1143 static int htab_save_later_pass(QEMUFile *f, sPAPREnvironment *spapr,
1144                                 int64_t max_ns)
1145 {
1146     bool final = max_ns < 0;
1147     int htabslots = HTAB_SIZE(spapr) / HASH_PTE_SIZE_64;
1148     int examined = 0, sent = 0;
1149     int index = spapr->htab_save_index;
1150     int64_t starttime = qemu_clock_get_ns(QEMU_CLOCK_REALTIME);
1151 
1152     assert(!spapr->htab_first_pass);
1153 
1154     do {
1155         int chunkstart, invalidstart;
1156 
1157         /* Consume non-dirty HPTEs */
1158         while ((index < htabslots)
1159                && !HPTE_DIRTY(HPTE(spapr->htab, index))) {
1160             index++;
1161             examined++;
1162         }
1163 
1164         chunkstart = index;
1165         /* Consume valid dirty HPTEs */
1166         while ((index < htabslots) && (index - chunkstart < USHRT_MAX)
1167                && HPTE_DIRTY(HPTE(spapr->htab, index))
1168                && HPTE_VALID(HPTE(spapr->htab, index))) {
1169             CLEAN_HPTE(HPTE(spapr->htab, index));
1170             index++;
1171             examined++;
1172         }
1173 
1174         invalidstart = index;
1175         /* Consume invalid dirty HPTEs */
1176         while ((index < htabslots) && (index - invalidstart < USHRT_MAX)
1177                && HPTE_DIRTY(HPTE(spapr->htab, index))
1178                && !HPTE_VALID(HPTE(spapr->htab, index))) {
1179             CLEAN_HPTE(HPTE(spapr->htab, index));
1180             index++;
1181             examined++;
1182         }
1183 
1184         if (index > chunkstart) {
1185             int n_valid = invalidstart - chunkstart;
1186             int n_invalid = index - invalidstart;
1187 
1188             qemu_put_be32(f, chunkstart);
1189             qemu_put_be16(f, n_valid);
1190             qemu_put_be16(f, n_invalid);
1191             qemu_put_buffer(f, HPTE(spapr->htab, chunkstart),
1192                             HASH_PTE_SIZE_64 * n_valid);
1193             sent += index - chunkstart;
1194 
1195             if (!final && (qemu_clock_get_ns(QEMU_CLOCK_REALTIME) - starttime) > max_ns) {
1196                 break;
1197             }
1198         }
1199 
1200         if (examined >= htabslots) {
1201             break;
1202         }
1203 
1204         if (index >= htabslots) {
1205             assert(index == htabslots);
1206             index = 0;
1207         }
1208     } while ((examined < htabslots) && (!qemu_file_rate_limit(f) || final));
1209 
1210     if (index >= htabslots) {
1211         assert(index == htabslots);
1212         index = 0;
1213     }
1214 
1215     spapr->htab_save_index = index;
1216 
1217     return (examined >= htabslots) && (sent == 0) ? 1 : 0;
1218 }
1219 
1220 #define MAX_ITERATION_NS    5000000 /* 5 ms */
1221 #define MAX_KVM_BUF_SIZE    2048
1222 
1223 static int htab_save_iterate(QEMUFile *f, void *opaque)
1224 {
1225     sPAPREnvironment *spapr = opaque;
1226     int rc = 0;
1227 
1228     /* Iteration header */
1229     qemu_put_be32(f, 0);
1230 
1231     if (!spapr->htab) {
1232         assert(kvm_enabled());
1233 
1234         rc = spapr_check_htab_fd(spapr);
1235         if (rc < 0) {
1236             return rc;
1237         }
1238 
1239         rc = kvmppc_save_htab(f, spapr->htab_fd,
1240                               MAX_KVM_BUF_SIZE, MAX_ITERATION_NS);
1241         if (rc < 0) {
1242             return rc;
1243         }
1244     } else  if (spapr->htab_first_pass) {
1245         htab_save_first_pass(f, spapr, MAX_ITERATION_NS);
1246     } else {
1247         rc = htab_save_later_pass(f, spapr, MAX_ITERATION_NS);
1248     }
1249 
1250     /* End marker */
1251     qemu_put_be32(f, 0);
1252     qemu_put_be16(f, 0);
1253     qemu_put_be16(f, 0);
1254 
1255     return rc;
1256 }
1257 
1258 static int htab_save_complete(QEMUFile *f, void *opaque)
1259 {
1260     sPAPREnvironment *spapr = opaque;
1261 
1262     /* Iteration header */
1263     qemu_put_be32(f, 0);
1264 
1265     if (!spapr->htab) {
1266         int rc;
1267 
1268         assert(kvm_enabled());
1269 
1270         rc = spapr_check_htab_fd(spapr);
1271         if (rc < 0) {
1272             return rc;
1273         }
1274 
1275         rc = kvmppc_save_htab(f, spapr->htab_fd, MAX_KVM_BUF_SIZE, -1);
1276         if (rc < 0) {
1277             return rc;
1278         }
1279         close(spapr->htab_fd);
1280         spapr->htab_fd = -1;
1281     } else {
1282         htab_save_later_pass(f, spapr, -1);
1283     }
1284 
1285     /* End marker */
1286     qemu_put_be32(f, 0);
1287     qemu_put_be16(f, 0);
1288     qemu_put_be16(f, 0);
1289 
1290     return 0;
1291 }
1292 
1293 static int htab_load(QEMUFile *f, void *opaque, int version_id)
1294 {
1295     sPAPREnvironment *spapr = opaque;
1296     uint32_t section_hdr;
1297     int fd = -1;
1298 
1299     if (version_id < 1 || version_id > 1) {
1300         fprintf(stderr, "htab_load() bad version\n");
1301         return -EINVAL;
1302     }
1303 
1304     section_hdr = qemu_get_be32(f);
1305 
1306     if (section_hdr) {
1307         /* First section, just the hash shift */
1308         if (spapr->htab_shift != section_hdr) {
1309             return -EINVAL;
1310         }
1311         return 0;
1312     }
1313 
1314     if (!spapr->htab) {
1315         assert(kvm_enabled());
1316 
1317         fd = kvmppc_get_htab_fd(true);
1318         if (fd < 0) {
1319             fprintf(stderr, "Unable to open fd to restore KVM hash table: %s\n",
1320                     strerror(errno));
1321         }
1322     }
1323 
1324     while (true) {
1325         uint32_t index;
1326         uint16_t n_valid, n_invalid;
1327 
1328         index = qemu_get_be32(f);
1329         n_valid = qemu_get_be16(f);
1330         n_invalid = qemu_get_be16(f);
1331 
1332         if ((index == 0) && (n_valid == 0) && (n_invalid == 0)) {
1333             /* End of Stream */
1334             break;
1335         }
1336 
1337         if ((index + n_valid + n_invalid) >
1338             (HTAB_SIZE(spapr) / HASH_PTE_SIZE_64)) {
1339             /* Bad index in stream */
1340             fprintf(stderr, "htab_load() bad index %d (%hd+%hd entries) "
1341                     "in htab stream (htab_shift=%d)\n", index, n_valid, n_invalid,
1342                     spapr->htab_shift);
1343             return -EINVAL;
1344         }
1345 
1346         if (spapr->htab) {
1347             if (n_valid) {
1348                 qemu_get_buffer(f, HPTE(spapr->htab, index),
1349                                 HASH_PTE_SIZE_64 * n_valid);
1350             }
1351             if (n_invalid) {
1352                 memset(HPTE(spapr->htab, index + n_valid), 0,
1353                        HASH_PTE_SIZE_64 * n_invalid);
1354             }
1355         } else {
1356             int rc;
1357 
1358             assert(fd >= 0);
1359 
1360             rc = kvmppc_load_htab_chunk(f, fd, index, n_valid, n_invalid);
1361             if (rc < 0) {
1362                 return rc;
1363             }
1364         }
1365     }
1366 
1367     if (!spapr->htab) {
1368         assert(fd >= 0);
1369         close(fd);
1370     }
1371 
1372     return 0;
1373 }
1374 
1375 static SaveVMHandlers savevm_htab_handlers = {
1376     .save_live_setup = htab_save_setup,
1377     .save_live_iterate = htab_save_iterate,
1378     .save_live_complete = htab_save_complete,
1379     .load_state = htab_load,
1380 };
1381 
1382 static void spapr_boot_set(void *opaque, const char *boot_device,
1383                            Error **errp)
1384 {
1385     MachineState *machine = MACHINE(qdev_get_machine());
1386     machine->boot_order = g_strdup(boot_device);
1387 }
1388 
1389 /* pSeries LPAR / sPAPR hardware init */
1390 static void ppc_spapr_init(MachineState *machine)
1391 {
1392     ram_addr_t ram_size = machine->ram_size;
1393     const char *cpu_model = machine->cpu_model;
1394     const char *kernel_filename = machine->kernel_filename;
1395     const char *kernel_cmdline = machine->kernel_cmdline;
1396     const char *initrd_filename = machine->initrd_filename;
1397     PowerPCCPU *cpu;
1398     CPUPPCState *env;
1399     PCIHostState *phb;
1400     int i;
1401     MemoryRegion *sysmem = get_system_memory();
1402     MemoryRegion *ram = g_new(MemoryRegion, 1);
1403     MemoryRegion *rma_region;
1404     void *rma = NULL;
1405     hwaddr rma_alloc_size;
1406     hwaddr node0_size = spapr_node0_size();
1407     uint32_t initrd_base = 0;
1408     long kernel_size = 0, initrd_size = 0;
1409     long load_limit, fw_size;
1410     bool kernel_le = false;
1411     char *filename;
1412 
1413     msi_supported = true;
1414 
1415     spapr = g_malloc0(sizeof(*spapr));
1416     QLIST_INIT(&spapr->phbs);
1417 
1418     cpu_ppc_hypercall = emulate_spapr_hypercall;
1419 
1420     /* Allocate RMA if necessary */
1421     rma_alloc_size = kvmppc_alloc_rma(&rma);
1422 
1423     if (rma_alloc_size == -1) {
1424         error_report("Unable to create RMA");
1425         exit(1);
1426     }
1427 
1428     if (rma_alloc_size && (rma_alloc_size < node0_size)) {
1429         spapr->rma_size = rma_alloc_size;
1430     } else {
1431         spapr->rma_size = node0_size;
1432 
1433         /* With KVM, we don't actually know whether KVM supports an
1434          * unbounded RMA (PR KVM) or is limited by the hash table size
1435          * (HV KVM using VRMA), so we always assume the latter
1436          *
1437          * In that case, we also limit the initial allocations for RTAS
1438          * etc... to 256M since we have no way to know what the VRMA size
1439          * is going to be as it depends on the size of the hash table
1440          * isn't determined yet.
1441          */
1442         if (kvm_enabled()) {
1443             spapr->vrma_adjust = 1;
1444             spapr->rma_size = MIN(spapr->rma_size, 0x10000000);
1445         }
1446     }
1447 
1448     if (spapr->rma_size > node0_size) {
1449         fprintf(stderr, "Error: Numa node 0 has to span the RMA (%#08"HWADDR_PRIx")\n",
1450                 spapr->rma_size);
1451         exit(1);
1452     }
1453 
1454     /* Setup a load limit for the ramdisk leaving room for SLOF and FDT */
1455     load_limit = MIN(spapr->rma_size, RTAS_MAX_ADDR) - FW_OVERHEAD;
1456 
1457     /* We aim for a hash table of size 1/128 the size of RAM.  The
1458      * normal rule of thumb is 1/64 the size of RAM, but that's much
1459      * more than needed for the Linux guests we support. */
1460     spapr->htab_shift = 18; /* Minimum architected size */
1461     while (spapr->htab_shift <= 46) {
1462         if ((1ULL << (spapr->htab_shift + 7)) >= ram_size) {
1463             break;
1464         }
1465         spapr->htab_shift++;
1466     }
1467 
1468     /* Set up Interrupt Controller before we create the VCPUs */
1469     spapr->icp = xics_system_init(machine,
1470                                   smp_cpus * kvmppc_smt_threads() / smp_threads,
1471                                   XICS_IRQS);
1472 
1473     /* init CPUs */
1474     if (cpu_model == NULL) {
1475         cpu_model = kvm_enabled() ? "host" : "POWER7";
1476     }
1477     for (i = 0; i < smp_cpus; i++) {
1478         cpu = cpu_ppc_init(cpu_model);
1479         if (cpu == NULL) {
1480             fprintf(stderr, "Unable to find PowerPC CPU definition\n");
1481             exit(1);
1482         }
1483         env = &cpu->env;
1484 
1485         /* Set time-base frequency to 512 MHz */
1486         cpu_ppc_tb_init(env, TIMEBASE_FREQ);
1487 
1488         /* PAPR always has exception vectors in RAM not ROM. To ensure this,
1489          * MSR[IP] should never be set.
1490          */
1491         env->msr_mask &= ~(1 << 6);
1492 
1493         /* Tell KVM that we're in PAPR mode */
1494         if (kvm_enabled()) {
1495             kvmppc_set_papr(cpu);
1496         }
1497 
1498         if (cpu->max_compat) {
1499             if (ppc_set_compat(cpu, cpu->max_compat) < 0) {
1500                 exit(1);
1501             }
1502         }
1503 
1504         xics_cpu_setup(spapr->icp, cpu);
1505 
1506         qemu_register_reset(spapr_cpu_reset, cpu);
1507     }
1508 
1509     if (kvm_enabled()) {
1510         /* Enable H_LOGICAL_CI_* so SLOF can talk to in-kernel devices */
1511         kvmppc_enable_logical_ci_hcalls();
1512     }
1513 
1514     /* allocate RAM */
1515     spapr->ram_limit = ram_size;
1516     memory_region_allocate_system_memory(ram, NULL, "ppc_spapr.ram",
1517                                          spapr->ram_limit);
1518     memory_region_add_subregion(sysmem, 0, ram);
1519 
1520     if (rma_alloc_size && rma) {
1521         rma_region = g_new(MemoryRegion, 1);
1522         memory_region_init_ram_ptr(rma_region, NULL, "ppc_spapr.rma",
1523                                    rma_alloc_size, rma);
1524         vmstate_register_ram_global(rma_region);
1525         memory_region_add_subregion(sysmem, 0, rma_region);
1526     }
1527 
1528     filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, "spapr-rtas.bin");
1529     if (!filename) {
1530         error_report("Could not find LPAR rtas '%s'", "spapr-rtas.bin");
1531         exit(1);
1532     }
1533     spapr->rtas_size = get_image_size(filename);
1534     spapr->rtas_blob = g_malloc(spapr->rtas_size);
1535     if (load_image_size(filename, spapr->rtas_blob, spapr->rtas_size) < 0) {
1536         error_report("Could not load LPAR rtas '%s'", filename);
1537         exit(1);
1538     }
1539     if (spapr->rtas_size > RTAS_MAX_SIZE) {
1540         error_report("RTAS too big ! 0x%zx bytes (max is 0x%x)",
1541                      (size_t)spapr->rtas_size, RTAS_MAX_SIZE);
1542         exit(1);
1543     }
1544     g_free(filename);
1545 
1546     /* Set up EPOW events infrastructure */
1547     spapr_events_init(spapr);
1548 
1549     /* Set up the RTC RTAS interfaces */
1550     spapr_rtc_create(spapr);
1551 
1552     /* Set up VIO bus */
1553     spapr->vio_bus = spapr_vio_bus_init();
1554 
1555     for (i = 0; i < MAX_SERIAL_PORTS; i++) {
1556         if (serial_hds[i]) {
1557             spapr_vty_create(spapr->vio_bus, serial_hds[i]);
1558         }
1559     }
1560 
1561     /* We always have at least the nvram device on VIO */
1562     spapr_create_nvram(spapr);
1563 
1564     /* Set up PCI */
1565     spapr_pci_rtas_init();
1566 
1567     phb = spapr_create_phb(spapr, 0);
1568 
1569     for (i = 0; i < nb_nics; i++) {
1570         NICInfo *nd = &nd_table[i];
1571 
1572         if (!nd->model) {
1573             nd->model = g_strdup("ibmveth");
1574         }
1575 
1576         if (strcmp(nd->model, "ibmveth") == 0) {
1577             spapr_vlan_create(spapr->vio_bus, nd);
1578         } else {
1579             pci_nic_init_nofail(&nd_table[i], phb->bus, nd->model, NULL);
1580         }
1581     }
1582 
1583     for (i = 0; i <= drive_get_max_bus(IF_SCSI); i++) {
1584         spapr_vscsi_create(spapr->vio_bus);
1585     }
1586 
1587     /* Graphics */
1588     if (spapr_vga_init(phb->bus)) {
1589         spapr->has_graphics = true;
1590         machine->usb |= defaults_enabled() && !machine->usb_disabled;
1591     }
1592 
1593     if (machine->usb) {
1594         pci_create_simple(phb->bus, -1, "pci-ohci");
1595 
1596         if (spapr->has_graphics) {
1597             USBBus *usb_bus = usb_bus_find(-1);
1598 
1599             usb_create_simple(usb_bus, "usb-kbd");
1600             usb_create_simple(usb_bus, "usb-mouse");
1601         }
1602     }
1603 
1604     if (spapr->rma_size < (MIN_RMA_SLOF << 20)) {
1605         fprintf(stderr, "qemu: pSeries SLOF firmware requires >= "
1606                 "%ldM guest RMA (Real Mode Area memory)\n", MIN_RMA_SLOF);
1607         exit(1);
1608     }
1609 
1610     if (kernel_filename) {
1611         uint64_t lowaddr = 0;
1612 
1613         kernel_size = load_elf(kernel_filename, translate_kernel_address, NULL,
1614                                NULL, &lowaddr, NULL, 1, ELF_MACHINE, 0);
1615         if (kernel_size == ELF_LOAD_WRONG_ENDIAN) {
1616             kernel_size = load_elf(kernel_filename,
1617                                    translate_kernel_address, NULL,
1618                                    NULL, &lowaddr, NULL, 0, ELF_MACHINE, 0);
1619             kernel_le = kernel_size > 0;
1620         }
1621         if (kernel_size < 0) {
1622             fprintf(stderr, "qemu: error loading %s: %s\n",
1623                     kernel_filename, load_elf_strerror(kernel_size));
1624             exit(1);
1625         }
1626 
1627         /* load initrd */
1628         if (initrd_filename) {
1629             /* Try to locate the initrd in the gap between the kernel
1630              * and the firmware. Add a bit of space just in case
1631              */
1632             initrd_base = (KERNEL_LOAD_ADDR + kernel_size + 0x1ffff) & ~0xffff;
1633             initrd_size = load_image_targphys(initrd_filename, initrd_base,
1634                                               load_limit - initrd_base);
1635             if (initrd_size < 0) {
1636                 fprintf(stderr, "qemu: could not load initial ram disk '%s'\n",
1637                         initrd_filename);
1638                 exit(1);
1639             }
1640         } else {
1641             initrd_base = 0;
1642             initrd_size = 0;
1643         }
1644     }
1645 
1646     if (bios_name == NULL) {
1647         bios_name = FW_FILE_NAME;
1648     }
1649     filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, bios_name);
1650     if (!filename) {
1651         error_report("Could not find LPAR firmware '%s'", bios_name);
1652         exit(1);
1653     }
1654     fw_size = load_image_targphys(filename, 0, FW_MAX_SIZE);
1655     if (fw_size <= 0) {
1656         error_report("Could not load LPAR firmware '%s'", filename);
1657         exit(1);
1658     }
1659     g_free(filename);
1660 
1661     spapr->entry_point = 0x100;
1662 
1663     vmstate_register(NULL, 0, &vmstate_spapr, spapr);
1664     register_savevm_live(NULL, "spapr/htab", -1, 1,
1665                          &savevm_htab_handlers, spapr);
1666 
1667     /* Prepare the device tree */
1668     spapr->fdt_skel = spapr_create_fdt_skel(initrd_base, initrd_size,
1669                                             kernel_size, kernel_le,
1670                                             kernel_cmdline,
1671                                             spapr->check_exception_irq);
1672     assert(spapr->fdt_skel != NULL);
1673 
1674     /* used by RTAS */
1675     QTAILQ_INIT(&spapr->ccs_list);
1676     qemu_register_reset(spapr_ccs_reset_hook, spapr);
1677 
1678     qemu_register_boot_set(spapr_boot_set, spapr);
1679 }
1680 
1681 static int spapr_kvm_type(const char *vm_type)
1682 {
1683     if (!vm_type) {
1684         return 0;
1685     }
1686 
1687     if (!strcmp(vm_type, "HV")) {
1688         return 1;
1689     }
1690 
1691     if (!strcmp(vm_type, "PR")) {
1692         return 2;
1693     }
1694 
1695     error_report("Unknown kvm-type specified '%s'", vm_type);
1696     exit(1);
1697 }
1698 
1699 /*
1700  * Implementation of an interface to adjust firmware path
1701  * for the bootindex property handling.
1702  */
1703 static char *spapr_get_fw_dev_path(FWPathProvider *p, BusState *bus,
1704                                    DeviceState *dev)
1705 {
1706 #define CAST(type, obj, name) \
1707     ((type *)object_dynamic_cast(OBJECT(obj), (name)))
1708     SCSIDevice *d = CAST(SCSIDevice,  dev, TYPE_SCSI_DEVICE);
1709     sPAPRPHBState *phb = CAST(sPAPRPHBState, dev, TYPE_SPAPR_PCI_HOST_BRIDGE);
1710 
1711     if (d) {
1712         void *spapr = CAST(void, bus->parent, "spapr-vscsi");
1713         VirtIOSCSI *virtio = CAST(VirtIOSCSI, bus->parent, TYPE_VIRTIO_SCSI);
1714         USBDevice *usb = CAST(USBDevice, bus->parent, TYPE_USB_DEVICE);
1715 
1716         if (spapr) {
1717             /*
1718              * Replace "channel@0/disk@0,0" with "disk@8000000000000000":
1719              * We use SRP luns of the form 8000 | (bus << 8) | (id << 5) | lun
1720              * in the top 16 bits of the 64-bit LUN
1721              */
1722             unsigned id = 0x8000 | (d->id << 8) | d->lun;
1723             return g_strdup_printf("%s@%"PRIX64, qdev_fw_name(dev),
1724                                    (uint64_t)id << 48);
1725         } else if (virtio) {
1726             /*
1727              * We use SRP luns of the form 01000000 | (target << 8) | lun
1728              * in the top 32 bits of the 64-bit LUN
1729              * Note: the quote above is from SLOF and it is wrong,
1730              * the actual binding is:
1731              * swap 0100 or 10 << or 20 << ( target lun-id -- srplun )
1732              */
1733             unsigned id = 0x1000000 | (d->id << 16) | d->lun;
1734             return g_strdup_printf("%s@%"PRIX64, qdev_fw_name(dev),
1735                                    (uint64_t)id << 32);
1736         } else if (usb) {
1737             /*
1738              * We use SRP luns of the form 01000000 | (usb-port << 16) | lun
1739              * in the top 32 bits of the 64-bit LUN
1740              */
1741             unsigned usb_port = atoi(usb->port->path);
1742             unsigned id = 0x1000000 | (usb_port << 16) | d->lun;
1743             return g_strdup_printf("%s@%"PRIX64, qdev_fw_name(dev),
1744                                    (uint64_t)id << 32);
1745         }
1746     }
1747 
1748     if (phb) {
1749         /* Replace "pci" with "pci@800000020000000" */
1750         return g_strdup_printf("pci@%"PRIX64, phb->buid);
1751     }
1752 
1753     return NULL;
1754 }
1755 
1756 static char *spapr_get_kvm_type(Object *obj, Error **errp)
1757 {
1758     sPAPRMachineState *sm = SPAPR_MACHINE(obj);
1759 
1760     return g_strdup(sm->kvm_type);
1761 }
1762 
1763 static void spapr_set_kvm_type(Object *obj, const char *value, Error **errp)
1764 {
1765     sPAPRMachineState *sm = SPAPR_MACHINE(obj);
1766 
1767     g_free(sm->kvm_type);
1768     sm->kvm_type = g_strdup(value);
1769 }
1770 
1771 static void spapr_machine_initfn(Object *obj)
1772 {
1773     object_property_add_str(obj, "kvm-type",
1774                             spapr_get_kvm_type, spapr_set_kvm_type, NULL);
1775     object_property_set_description(obj, "kvm-type",
1776                                     "Specifies the KVM virtualization mode (HV, PR)",
1777                                     NULL);
1778 }
1779 
1780 static void ppc_cpu_do_nmi_on_cpu(void *arg)
1781 {
1782     CPUState *cs = arg;
1783 
1784     cpu_synchronize_state(cs);
1785     ppc_cpu_do_system_reset(cs);
1786 }
1787 
1788 static void spapr_nmi(NMIState *n, int cpu_index, Error **errp)
1789 {
1790     CPUState *cs;
1791 
1792     CPU_FOREACH(cs) {
1793         async_run_on_cpu(cs, ppc_cpu_do_nmi_on_cpu, cs);
1794     }
1795 }
1796 
1797 static void spapr_machine_class_init(ObjectClass *oc, void *data)
1798 {
1799     MachineClass *mc = MACHINE_CLASS(oc);
1800     FWPathProviderClass *fwc = FW_PATH_PROVIDER_CLASS(oc);
1801     NMIClass *nc = NMI_CLASS(oc);
1802 
1803     mc->init = ppc_spapr_init;
1804     mc->reset = ppc_spapr_reset;
1805     mc->block_default_type = IF_SCSI;
1806     mc->max_cpus = MAX_CPUS;
1807     mc->no_parallel = 1;
1808     mc->default_boot_order = "";
1809     mc->default_ram_size = 512 * M_BYTE;
1810     mc->kvm_type = spapr_kvm_type;
1811     mc->has_dynamic_sysbus = true;
1812 
1813     fwc->get_dev_path = spapr_get_fw_dev_path;
1814     nc->nmi_monitor_handler = spapr_nmi;
1815 }
1816 
1817 static const TypeInfo spapr_machine_info = {
1818     .name          = TYPE_SPAPR_MACHINE,
1819     .parent        = TYPE_MACHINE,
1820     .abstract      = true,
1821     .instance_size = sizeof(sPAPRMachineState),
1822     .instance_init = spapr_machine_initfn,
1823     .class_init    = spapr_machine_class_init,
1824     .interfaces = (InterfaceInfo[]) {
1825         { TYPE_FW_PATH_PROVIDER },
1826         { TYPE_NMI },
1827         { }
1828     },
1829 };
1830 
1831 #define SPAPR_COMPAT_2_3 \
1832         HW_COMPAT_2_3 \
1833         {\
1834             .driver   = "spapr-pci-host-bridge",\
1835             .property = "dynamic-reconfiguration",\
1836             .value    = "off",\
1837         },
1838 
1839 #define SPAPR_COMPAT_2_2 \
1840         SPAPR_COMPAT_2_3 \
1841         HW_COMPAT_2_2 \
1842         {\
1843             .driver   = TYPE_SPAPR_PCI_HOST_BRIDGE,\
1844             .property = "mem_win_size",\
1845             .value    = "0x20000000",\
1846         },
1847 
1848 #define SPAPR_COMPAT_2_1 \
1849         SPAPR_COMPAT_2_2 \
1850         HW_COMPAT_2_1
1851 
1852 static void spapr_compat_2_3(Object *obj)
1853 {
1854 }
1855 
1856 static void spapr_compat_2_2(Object *obj)
1857 {
1858     spapr_compat_2_3(obj);
1859 }
1860 
1861 static void spapr_compat_2_1(Object *obj)
1862 {
1863     spapr_compat_2_2(obj);
1864 }
1865 
1866 static void spapr_machine_2_3_instance_init(Object *obj)
1867 {
1868     spapr_compat_2_3(obj);
1869     spapr_machine_initfn(obj);
1870 }
1871 
1872 static void spapr_machine_2_2_instance_init(Object *obj)
1873 {
1874     spapr_compat_2_2(obj);
1875     spapr_machine_initfn(obj);
1876 }
1877 
1878 static void spapr_machine_2_1_instance_init(Object *obj)
1879 {
1880     spapr_compat_2_1(obj);
1881     spapr_machine_initfn(obj);
1882 }
1883 
1884 static void spapr_machine_2_1_class_init(ObjectClass *oc, void *data)
1885 {
1886     MachineClass *mc = MACHINE_CLASS(oc);
1887     static GlobalProperty compat_props[] = {
1888         SPAPR_COMPAT_2_1
1889         { /* end of list */ }
1890     };
1891 
1892     mc->name = "pseries-2.1";
1893     mc->desc = "pSeries Logical Partition (PAPR compliant) v2.1";
1894     mc->compat_props = compat_props;
1895 }
1896 
1897 static const TypeInfo spapr_machine_2_1_info = {
1898     .name          = TYPE_SPAPR_MACHINE "2.1",
1899     .parent        = TYPE_SPAPR_MACHINE,
1900     .class_init    = spapr_machine_2_1_class_init,
1901     .instance_init = spapr_machine_2_1_instance_init,
1902 };
1903 
1904 static void spapr_machine_2_2_class_init(ObjectClass *oc, void *data)
1905 {
1906     static GlobalProperty compat_props[] = {
1907         SPAPR_COMPAT_2_2
1908         { /* end of list */ }
1909     };
1910     MachineClass *mc = MACHINE_CLASS(oc);
1911 
1912     mc->name = "pseries-2.2";
1913     mc->desc = "pSeries Logical Partition (PAPR compliant) v2.2";
1914     mc->compat_props = compat_props;
1915 }
1916 
1917 static const TypeInfo spapr_machine_2_2_info = {
1918     .name          = TYPE_SPAPR_MACHINE "2.2",
1919     .parent        = TYPE_SPAPR_MACHINE,
1920     .class_init    = spapr_machine_2_2_class_init,
1921     .instance_init = spapr_machine_2_2_instance_init,
1922 };
1923 
1924 static void spapr_machine_2_3_class_init(ObjectClass *oc, void *data)
1925 {
1926     static GlobalProperty compat_props[] = {
1927         SPAPR_COMPAT_2_3
1928         { /* end of list */ }
1929     };
1930     MachineClass *mc = MACHINE_CLASS(oc);
1931 
1932     mc->name = "pseries-2.3";
1933     mc->desc = "pSeries Logical Partition (PAPR compliant) v2.3";
1934     mc->compat_props = compat_props;
1935 }
1936 
1937 static const TypeInfo spapr_machine_2_3_info = {
1938     .name          = TYPE_SPAPR_MACHINE "2.3",
1939     .parent        = TYPE_SPAPR_MACHINE,
1940     .class_init    = spapr_machine_2_3_class_init,
1941     .instance_init = spapr_machine_2_3_instance_init,
1942 };
1943 
1944 static void spapr_machine_2_4_class_init(ObjectClass *oc, void *data)
1945 {
1946     MachineClass *mc = MACHINE_CLASS(oc);
1947 
1948     mc->name = "pseries-2.4";
1949     mc->desc = "pSeries Logical Partition (PAPR compliant) v2.4";
1950     mc->alias = "pseries";
1951     mc->is_default = 1;
1952 }
1953 
1954 static const TypeInfo spapr_machine_2_4_info = {
1955     .name          = TYPE_SPAPR_MACHINE "2.4",
1956     .parent        = TYPE_SPAPR_MACHINE,
1957     .class_init    = spapr_machine_2_4_class_init,
1958 };
1959 
1960 static void spapr_machine_register_types(void)
1961 {
1962     type_register_static(&spapr_machine_info);
1963     type_register_static(&spapr_machine_2_1_info);
1964     type_register_static(&spapr_machine_2_2_info);
1965     type_register_static(&spapr_machine_2_3_info);
1966     type_register_static(&spapr_machine_2_4_info);
1967 }
1968 
1969 type_init(spapr_machine_register_types)
1970