xref: /openbmc/qemu/hw/ppc/spapr.c (revision db4ef288f4a6d285b39dc8ac477092d76971a300)
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 static XICSState *try_create_xics(const char *type, int nr_servers,
94                                   int nr_irqs, Error **errp)
95 {
96     Error *err = NULL;
97     DeviceState *dev;
98 
99     dev = qdev_create(NULL, type);
100     qdev_prop_set_uint32(dev, "nr_servers", nr_servers);
101     qdev_prop_set_uint32(dev, "nr_irqs", nr_irqs);
102     object_property_set_bool(OBJECT(dev), true, "realized", &err);
103     if (err) {
104         error_propagate(errp, err);
105         object_unparent(OBJECT(dev));
106         return NULL;
107     }
108     return XICS_COMMON(dev);
109 }
110 
111 static XICSState *xics_system_init(MachineState *machine,
112                                    int nr_servers, int nr_irqs)
113 {
114     XICSState *icp = NULL;
115 
116     if (kvm_enabled()) {
117         Error *err = NULL;
118 
119         if (machine_kernel_irqchip_allowed(machine)) {
120             icp = try_create_xics(TYPE_KVM_XICS, nr_servers, nr_irqs, &err);
121         }
122         if (machine_kernel_irqchip_required(machine) && !icp) {
123             error_report("kernel_irqchip requested but unavailable: %s",
124                          error_get_pretty(err));
125         }
126     }
127 
128     if (!icp) {
129         icp = try_create_xics(TYPE_XICS, nr_servers, nr_irqs, &error_abort);
130     }
131 
132     return icp;
133 }
134 
135 static int spapr_fixup_cpu_smt_dt(void *fdt, int offset, PowerPCCPU *cpu,
136                                   int smt_threads)
137 {
138     int i, ret = 0;
139     uint32_t servers_prop[smt_threads];
140     uint32_t gservers_prop[smt_threads * 2];
141     int index = ppc_get_vcpu_dt_id(cpu);
142 
143     if (cpu->cpu_version) {
144         ret = fdt_setprop_cell(fdt, offset, "cpu-version", cpu->cpu_version);
145         if (ret < 0) {
146             return ret;
147         }
148     }
149 
150     /* Build interrupt servers and gservers properties */
151     for (i = 0; i < smt_threads; i++) {
152         servers_prop[i] = cpu_to_be32(index + i);
153         /* Hack, direct the group queues back to cpu 0 */
154         gservers_prop[i*2] = cpu_to_be32(index + i);
155         gservers_prop[i*2 + 1] = 0;
156     }
157     ret = fdt_setprop(fdt, offset, "ibm,ppc-interrupt-server#s",
158                       servers_prop, sizeof(servers_prop));
159     if (ret < 0) {
160         return ret;
161     }
162     ret = fdt_setprop(fdt, offset, "ibm,ppc-interrupt-gserver#s",
163                       gservers_prop, sizeof(gservers_prop));
164 
165     return ret;
166 }
167 
168 static int spapr_fixup_cpu_dt(void *fdt, sPAPRMachineState *spapr)
169 {
170     int ret = 0, offset, cpus_offset;
171     CPUState *cs;
172     char cpu_model[32];
173     int smt = kvmppc_smt_threads();
174     uint32_t pft_size_prop[] = {0, cpu_to_be32(spapr->htab_shift)};
175 
176     CPU_FOREACH(cs) {
177         PowerPCCPU *cpu = POWERPC_CPU(cs);
178         DeviceClass *dc = DEVICE_GET_CLASS(cs);
179         int index = ppc_get_vcpu_dt_id(cpu);
180         uint32_t associativity[] = {cpu_to_be32(0x5),
181                                     cpu_to_be32(0x0),
182                                     cpu_to_be32(0x0),
183                                     cpu_to_be32(0x0),
184                                     cpu_to_be32(cs->numa_node),
185                                     cpu_to_be32(index)};
186 
187         if ((index % smt) != 0) {
188             continue;
189         }
190 
191         snprintf(cpu_model, 32, "%s@%x", dc->fw_name, index);
192 
193         cpus_offset = fdt_path_offset(fdt, "/cpus");
194         if (cpus_offset < 0) {
195             cpus_offset = fdt_add_subnode(fdt, fdt_path_offset(fdt, "/"),
196                                           "cpus");
197             if (cpus_offset < 0) {
198                 return cpus_offset;
199             }
200         }
201         offset = fdt_subnode_offset(fdt, cpus_offset, cpu_model);
202         if (offset < 0) {
203             offset = fdt_add_subnode(fdt, cpus_offset, cpu_model);
204             if (offset < 0) {
205                 return offset;
206             }
207         }
208 
209         if (nb_numa_nodes > 1) {
210             ret = fdt_setprop(fdt, offset, "ibm,associativity", associativity,
211                               sizeof(associativity));
212             if (ret < 0) {
213                 return ret;
214             }
215         }
216 
217         ret = fdt_setprop(fdt, offset, "ibm,pft-size",
218                           pft_size_prop, sizeof(pft_size_prop));
219         if (ret < 0) {
220             return ret;
221         }
222 
223         ret = spapr_fixup_cpu_smt_dt(fdt, offset, cpu,
224                                      ppc_get_compat_smt_threads(cpu));
225         if (ret < 0) {
226             return ret;
227         }
228     }
229     return ret;
230 }
231 
232 
233 static size_t create_page_sizes_prop(CPUPPCState *env, uint32_t *prop,
234                                      size_t maxsize)
235 {
236     size_t maxcells = maxsize / sizeof(uint32_t);
237     int i, j, count;
238     uint32_t *p = prop;
239 
240     for (i = 0; i < PPC_PAGE_SIZES_MAX_SZ; i++) {
241         struct ppc_one_seg_page_size *sps = &env->sps.sps[i];
242 
243         if (!sps->page_shift) {
244             break;
245         }
246         for (count = 0; count < PPC_PAGE_SIZES_MAX_SZ; count++) {
247             if (sps->enc[count].page_shift == 0) {
248                 break;
249             }
250         }
251         if ((p - prop) >= (maxcells - 3 - count * 2)) {
252             break;
253         }
254         *(p++) = cpu_to_be32(sps->page_shift);
255         *(p++) = cpu_to_be32(sps->slb_enc);
256         *(p++) = cpu_to_be32(count);
257         for (j = 0; j < count; j++) {
258             *(p++) = cpu_to_be32(sps->enc[j].page_shift);
259             *(p++) = cpu_to_be32(sps->enc[j].pte_enc);
260         }
261     }
262 
263     return (p - prop) * sizeof(uint32_t);
264 }
265 
266 static hwaddr spapr_node0_size(void)
267 {
268     MachineState *machine = MACHINE(qdev_get_machine());
269 
270     if (nb_numa_nodes) {
271         int i;
272         for (i = 0; i < nb_numa_nodes; ++i) {
273             if (numa_info[i].node_mem) {
274                 return MIN(pow2floor(numa_info[i].node_mem),
275                            machine->ram_size);
276             }
277         }
278     }
279     return machine->ram_size;
280 }
281 
282 #define _FDT(exp) \
283     do { \
284         int ret = (exp);                                           \
285         if (ret < 0) {                                             \
286             fprintf(stderr, "qemu: error creating device tree: %s: %s\n", \
287                     #exp, fdt_strerror(ret));                      \
288             exit(1);                                               \
289         }                                                          \
290     } while (0)
291 
292 static void add_str(GString *s, const gchar *s1)
293 {
294     g_string_append_len(s, s1, strlen(s1) + 1);
295 }
296 
297 static void *spapr_create_fdt_skel(hwaddr initrd_base,
298                                    hwaddr initrd_size,
299                                    hwaddr kernel_size,
300                                    bool little_endian,
301                                    const char *kernel_cmdline,
302                                    uint32_t epow_irq)
303 {
304     void *fdt;
305     CPUState *cs;
306     uint32_t start_prop = cpu_to_be32(initrd_base);
307     uint32_t end_prop = cpu_to_be32(initrd_base + initrd_size);
308     GString *hypertas = g_string_sized_new(256);
309     GString *qemu_hypertas = g_string_sized_new(256);
310     uint32_t refpoints[] = {cpu_to_be32(0x4), cpu_to_be32(0x4)};
311     uint32_t interrupt_server_ranges_prop[] = {0, cpu_to_be32(max_cpus)};
312     int smt = kvmppc_smt_threads();
313     unsigned char vec5[] = {0x0, 0x0, 0x0, 0x0, 0x0, 0x80};
314     QemuOpts *opts = qemu_opts_find(qemu_find_opts("smp-opts"), NULL);
315     unsigned sockets = opts ? qemu_opt_get_number(opts, "sockets", 0) : 0;
316     uint32_t cpus_per_socket = sockets ? (smp_cpus / sockets) : 1;
317     char *buf;
318 
319     add_str(hypertas, "hcall-pft");
320     add_str(hypertas, "hcall-term");
321     add_str(hypertas, "hcall-dabr");
322     add_str(hypertas, "hcall-interrupt");
323     add_str(hypertas, "hcall-tce");
324     add_str(hypertas, "hcall-vio");
325     add_str(hypertas, "hcall-splpar");
326     add_str(hypertas, "hcall-bulk");
327     add_str(hypertas, "hcall-set-mode");
328     add_str(qemu_hypertas, "hcall-memop1");
329 
330     fdt = g_malloc0(FDT_MAX_SIZE);
331     _FDT((fdt_create(fdt, FDT_MAX_SIZE)));
332 
333     if (kernel_size) {
334         _FDT((fdt_add_reservemap_entry(fdt, KERNEL_LOAD_ADDR, kernel_size)));
335     }
336     if (initrd_size) {
337         _FDT((fdt_add_reservemap_entry(fdt, initrd_base, initrd_size)));
338     }
339     _FDT((fdt_finish_reservemap(fdt)));
340 
341     /* Root node */
342     _FDT((fdt_begin_node(fdt, "")));
343     _FDT((fdt_property_string(fdt, "device_type", "chrp")));
344     _FDT((fdt_property_string(fdt, "model", "IBM pSeries (emulated by qemu)")));
345     _FDT((fdt_property_string(fdt, "compatible", "qemu,pseries")));
346 
347     /*
348      * Add info to guest to indentify which host is it being run on
349      * and what is the uuid of the guest
350      */
351     if (kvmppc_get_host_model(&buf)) {
352         _FDT((fdt_property_string(fdt, "host-model", buf)));
353         g_free(buf);
354     }
355     if (kvmppc_get_host_serial(&buf)) {
356         _FDT((fdt_property_string(fdt, "host-serial", buf)));
357         g_free(buf);
358     }
359 
360     buf = g_strdup_printf(UUID_FMT, qemu_uuid[0], qemu_uuid[1],
361                           qemu_uuid[2], qemu_uuid[3], qemu_uuid[4],
362                           qemu_uuid[5], qemu_uuid[6], qemu_uuid[7],
363                           qemu_uuid[8], qemu_uuid[9], qemu_uuid[10],
364                           qemu_uuid[11], qemu_uuid[12], qemu_uuid[13],
365                           qemu_uuid[14], qemu_uuid[15]);
366 
367     _FDT((fdt_property_string(fdt, "vm,uuid", buf)));
368     g_free(buf);
369 
370     _FDT((fdt_property_cell(fdt, "#address-cells", 0x2)));
371     _FDT((fdt_property_cell(fdt, "#size-cells", 0x2)));
372 
373     /* /chosen */
374     _FDT((fdt_begin_node(fdt, "chosen")));
375 
376     /* Set Form1_affinity */
377     _FDT((fdt_property(fdt, "ibm,architecture-vec-5", vec5, sizeof(vec5))));
378 
379     _FDT((fdt_property_string(fdt, "bootargs", kernel_cmdline)));
380     _FDT((fdt_property(fdt, "linux,initrd-start",
381                        &start_prop, sizeof(start_prop))));
382     _FDT((fdt_property(fdt, "linux,initrd-end",
383                        &end_prop, sizeof(end_prop))));
384     if (kernel_size) {
385         uint64_t kprop[2] = { cpu_to_be64(KERNEL_LOAD_ADDR),
386                               cpu_to_be64(kernel_size) };
387 
388         _FDT((fdt_property(fdt, "qemu,boot-kernel", &kprop, sizeof(kprop))));
389         if (little_endian) {
390             _FDT((fdt_property(fdt, "qemu,boot-kernel-le", NULL, 0)));
391         }
392     }
393     if (boot_menu) {
394         _FDT((fdt_property_cell(fdt, "qemu,boot-menu", boot_menu)));
395     }
396     _FDT((fdt_property_cell(fdt, "qemu,graphic-width", graphic_width)));
397     _FDT((fdt_property_cell(fdt, "qemu,graphic-height", graphic_height)));
398     _FDT((fdt_property_cell(fdt, "qemu,graphic-depth", graphic_depth)));
399 
400     _FDT((fdt_end_node(fdt)));
401 
402     /* cpus */
403     _FDT((fdt_begin_node(fdt, "cpus")));
404 
405     _FDT((fdt_property_cell(fdt, "#address-cells", 0x1)));
406     _FDT((fdt_property_cell(fdt, "#size-cells", 0x0)));
407 
408     CPU_FOREACH(cs) {
409         PowerPCCPU *cpu = POWERPC_CPU(cs);
410         CPUPPCState *env = &cpu->env;
411         DeviceClass *dc = DEVICE_GET_CLASS(cs);
412         PowerPCCPUClass *pcc = POWERPC_CPU_GET_CLASS(cs);
413         int index = ppc_get_vcpu_dt_id(cpu);
414         char *nodename;
415         uint32_t segs[] = {cpu_to_be32(28), cpu_to_be32(40),
416                            0xffffffff, 0xffffffff};
417         uint32_t tbfreq = kvm_enabled() ? kvmppc_get_tbfreq() : TIMEBASE_FREQ;
418         uint32_t cpufreq = kvm_enabled() ? kvmppc_get_clockfreq() : 1000000000;
419         uint32_t page_sizes_prop[64];
420         size_t page_sizes_prop_size;
421 
422         if ((index % smt) != 0) {
423             continue;
424         }
425 
426         nodename = g_strdup_printf("%s@%x", dc->fw_name, index);
427 
428         _FDT((fdt_begin_node(fdt, nodename)));
429 
430         g_free(nodename);
431 
432         _FDT((fdt_property_cell(fdt, "reg", index)));
433         _FDT((fdt_property_string(fdt, "device_type", "cpu")));
434 
435         _FDT((fdt_property_cell(fdt, "cpu-version", env->spr[SPR_PVR])));
436         _FDT((fdt_property_cell(fdt, "d-cache-block-size",
437                                 env->dcache_line_size)));
438         _FDT((fdt_property_cell(fdt, "d-cache-line-size",
439                                 env->dcache_line_size)));
440         _FDT((fdt_property_cell(fdt, "i-cache-block-size",
441                                 env->icache_line_size)));
442         _FDT((fdt_property_cell(fdt, "i-cache-line-size",
443                                 env->icache_line_size)));
444 
445         if (pcc->l1_dcache_size) {
446             _FDT((fdt_property_cell(fdt, "d-cache-size", pcc->l1_dcache_size)));
447         } else {
448             fprintf(stderr, "Warning: Unknown L1 dcache size for cpu\n");
449         }
450         if (pcc->l1_icache_size) {
451             _FDT((fdt_property_cell(fdt, "i-cache-size", pcc->l1_icache_size)));
452         } else {
453             fprintf(stderr, "Warning: Unknown L1 icache size for cpu\n");
454         }
455 
456         _FDT((fdt_property_cell(fdt, "timebase-frequency", tbfreq)));
457         _FDT((fdt_property_cell(fdt, "clock-frequency", cpufreq)));
458         _FDT((fdt_property_cell(fdt, "ibm,slb-size", env->slb_nr)));
459         _FDT((fdt_property_string(fdt, "status", "okay")));
460         _FDT((fdt_property(fdt, "64-bit", NULL, 0)));
461 
462         if (env->spr_cb[SPR_PURR].oea_read) {
463             _FDT((fdt_property(fdt, "ibm,purr", NULL, 0)));
464         }
465 
466         if (env->mmu_model & POWERPC_MMU_1TSEG) {
467             _FDT((fdt_property(fdt, "ibm,processor-segment-sizes",
468                                segs, sizeof(segs))));
469         }
470 
471         /* Advertise VMX/VSX (vector extensions) if available
472          *   0 / no property == no vector extensions
473          *   1               == VMX / Altivec available
474          *   2               == VSX available */
475         if (env->insns_flags & PPC_ALTIVEC) {
476             uint32_t vmx = (env->insns_flags2 & PPC2_VSX) ? 2 : 1;
477 
478             _FDT((fdt_property_cell(fdt, "ibm,vmx", vmx)));
479         }
480 
481         /* Advertise DFP (Decimal Floating Point) if available
482          *   0 / no property == no DFP
483          *   1               == DFP available */
484         if (env->insns_flags2 & PPC2_DFP) {
485             _FDT((fdt_property_cell(fdt, "ibm,dfp", 1)));
486         }
487 
488         page_sizes_prop_size = create_page_sizes_prop(env, page_sizes_prop,
489                                                       sizeof(page_sizes_prop));
490         if (page_sizes_prop_size) {
491             _FDT((fdt_property(fdt, "ibm,segment-page-sizes",
492                                page_sizes_prop, page_sizes_prop_size)));
493         }
494 
495         _FDT((fdt_property_cell(fdt, "ibm,chip-id",
496                                 cs->cpu_index / cpus_per_socket)));
497 
498         _FDT((fdt_end_node(fdt)));
499     }
500 
501     _FDT((fdt_end_node(fdt)));
502 
503     /* RTAS */
504     _FDT((fdt_begin_node(fdt, "rtas")));
505 
506     if (!kvm_enabled() || kvmppc_spapr_use_multitce()) {
507         add_str(hypertas, "hcall-multi-tce");
508     }
509     _FDT((fdt_property(fdt, "ibm,hypertas-functions", hypertas->str,
510                        hypertas->len)));
511     g_string_free(hypertas, TRUE);
512     _FDT((fdt_property(fdt, "qemu,hypertas-functions", qemu_hypertas->str,
513                        qemu_hypertas->len)));
514     g_string_free(qemu_hypertas, TRUE);
515 
516     _FDT((fdt_property(fdt, "ibm,associativity-reference-points",
517         refpoints, sizeof(refpoints))));
518 
519     _FDT((fdt_property_cell(fdt, "rtas-error-log-max", RTAS_ERROR_LOG_MAX)));
520     _FDT((fdt_property_cell(fdt, "rtas-event-scan-rate",
521                             RTAS_EVENT_SCAN_RATE)));
522 
523     /*
524      * According to PAPR, rtas ibm,os-term does not guarantee a return
525      * back to the guest cpu.
526      *
527      * While an additional ibm,extended-os-term property indicates that
528      * rtas call return will always occur. Set this property.
529      */
530     _FDT((fdt_property(fdt, "ibm,extended-os-term", NULL, 0)));
531 
532     _FDT((fdt_end_node(fdt)));
533 
534     /* interrupt controller */
535     _FDT((fdt_begin_node(fdt, "interrupt-controller")));
536 
537     _FDT((fdt_property_string(fdt, "device_type",
538                               "PowerPC-External-Interrupt-Presentation")));
539     _FDT((fdt_property_string(fdt, "compatible", "IBM,ppc-xicp")));
540     _FDT((fdt_property(fdt, "interrupt-controller", NULL, 0)));
541     _FDT((fdt_property(fdt, "ibm,interrupt-server-ranges",
542                        interrupt_server_ranges_prop,
543                        sizeof(interrupt_server_ranges_prop))));
544     _FDT((fdt_property_cell(fdt, "#interrupt-cells", 2)));
545     _FDT((fdt_property_cell(fdt, "linux,phandle", PHANDLE_XICP)));
546     _FDT((fdt_property_cell(fdt, "phandle", PHANDLE_XICP)));
547 
548     _FDT((fdt_end_node(fdt)));
549 
550     /* vdevice */
551     _FDT((fdt_begin_node(fdt, "vdevice")));
552 
553     _FDT((fdt_property_string(fdt, "device_type", "vdevice")));
554     _FDT((fdt_property_string(fdt, "compatible", "IBM,vdevice")));
555     _FDT((fdt_property_cell(fdt, "#address-cells", 0x1)));
556     _FDT((fdt_property_cell(fdt, "#size-cells", 0x0)));
557     _FDT((fdt_property_cell(fdt, "#interrupt-cells", 0x2)));
558     _FDT((fdt_property(fdt, "interrupt-controller", NULL, 0)));
559 
560     _FDT((fdt_end_node(fdt)));
561 
562     /* event-sources */
563     spapr_events_fdt_skel(fdt, epow_irq);
564 
565     /* /hypervisor node */
566     if (kvm_enabled()) {
567         uint8_t hypercall[16];
568 
569         /* indicate KVM hypercall interface */
570         _FDT((fdt_begin_node(fdt, "hypervisor")));
571         _FDT((fdt_property_string(fdt, "compatible", "linux,kvm")));
572         if (kvmppc_has_cap_fixup_hcalls()) {
573             /*
574              * Older KVM versions with older guest kernels were broken with the
575              * magic page, don't allow the guest to map it.
576              */
577             kvmppc_get_hypercall(first_cpu->env_ptr, hypercall,
578                                  sizeof(hypercall));
579             _FDT((fdt_property(fdt, "hcall-instructions", hypercall,
580                               sizeof(hypercall))));
581         }
582         _FDT((fdt_end_node(fdt)));
583     }
584 
585     _FDT((fdt_end_node(fdt))); /* close root node */
586     _FDT((fdt_finish(fdt)));
587 
588     return fdt;
589 }
590 
591 int spapr_h_cas_compose_response(sPAPRMachineState *spapr,
592                                  target_ulong addr, target_ulong size)
593 {
594     void *fdt, *fdt_skel;
595     sPAPRDeviceTreeUpdateHeader hdr = { .version_id = 1 };
596 
597     size -= sizeof(hdr);
598 
599     /* Create sceleton */
600     fdt_skel = g_malloc0(size);
601     _FDT((fdt_create(fdt_skel, size)));
602     _FDT((fdt_begin_node(fdt_skel, "")));
603     _FDT((fdt_end_node(fdt_skel)));
604     _FDT((fdt_finish(fdt_skel)));
605     fdt = g_malloc0(size);
606     _FDT((fdt_open_into(fdt_skel, fdt, size)));
607     g_free(fdt_skel);
608 
609     /* Fix skeleton up */
610     _FDT((spapr_fixup_cpu_dt(fdt, spapr)));
611 
612     /* Pack resulting tree */
613     _FDT((fdt_pack(fdt)));
614 
615     if (fdt_totalsize(fdt) + sizeof(hdr) > size) {
616         trace_spapr_cas_failed(size);
617         return -1;
618     }
619 
620     cpu_physical_memory_write(addr, &hdr, sizeof(hdr));
621     cpu_physical_memory_write(addr + sizeof(hdr), fdt, fdt_totalsize(fdt));
622     trace_spapr_cas_continue(fdt_totalsize(fdt) + sizeof(hdr));
623     g_free(fdt);
624 
625     return 0;
626 }
627 
628 static void spapr_populate_memory_node(void *fdt, int nodeid, hwaddr start,
629                                        hwaddr size)
630 {
631     uint32_t associativity[] = {
632         cpu_to_be32(0x4), /* length */
633         cpu_to_be32(0x0), cpu_to_be32(0x0),
634         cpu_to_be32(0x0), cpu_to_be32(nodeid)
635     };
636     char mem_name[32];
637     uint64_t mem_reg_property[2];
638     int off;
639 
640     mem_reg_property[0] = cpu_to_be64(start);
641     mem_reg_property[1] = cpu_to_be64(size);
642 
643     sprintf(mem_name, "memory@" TARGET_FMT_lx, start);
644     off = fdt_add_subnode(fdt, 0, mem_name);
645     _FDT(off);
646     _FDT((fdt_setprop_string(fdt, off, "device_type", "memory")));
647     _FDT((fdt_setprop(fdt, off, "reg", mem_reg_property,
648                       sizeof(mem_reg_property))));
649     _FDT((fdt_setprop(fdt, off, "ibm,associativity", associativity,
650                       sizeof(associativity))));
651 }
652 
653 static int spapr_populate_memory(sPAPRMachineState *spapr, void *fdt)
654 {
655     MachineState *machine = MACHINE(spapr);
656     hwaddr mem_start, node_size;
657     int i, nb_nodes = nb_numa_nodes;
658     NodeInfo *nodes = numa_info;
659     NodeInfo ramnode;
660 
661     /* No NUMA nodes, assume there is just one node with whole RAM */
662     if (!nb_numa_nodes) {
663         nb_nodes = 1;
664         ramnode.node_mem = machine->ram_size;
665         nodes = &ramnode;
666     }
667 
668     for (i = 0, mem_start = 0; i < nb_nodes; ++i) {
669         if (!nodes[i].node_mem) {
670             continue;
671         }
672         if (mem_start >= machine->ram_size) {
673             node_size = 0;
674         } else {
675             node_size = nodes[i].node_mem;
676             if (node_size > machine->ram_size - mem_start) {
677                 node_size = machine->ram_size - mem_start;
678             }
679         }
680         if (!mem_start) {
681             /* ppc_spapr_init() checks for rma_size <= node0_size already */
682             spapr_populate_memory_node(fdt, i, 0, spapr->rma_size);
683             mem_start += spapr->rma_size;
684             node_size -= spapr->rma_size;
685         }
686         for ( ; node_size; ) {
687             hwaddr sizetmp = pow2floor(node_size);
688 
689             /* mem_start != 0 here */
690             if (ctzl(mem_start) < ctzl(sizetmp)) {
691                 sizetmp = 1ULL << ctzl(mem_start);
692             }
693 
694             spapr_populate_memory_node(fdt, i, mem_start, sizetmp);
695             node_size -= sizetmp;
696             mem_start += sizetmp;
697         }
698     }
699 
700     return 0;
701 }
702 
703 static void spapr_finalize_fdt(sPAPRMachineState *spapr,
704                                hwaddr fdt_addr,
705                                hwaddr rtas_addr,
706                                hwaddr rtas_size)
707 {
708     MachineState *machine = MACHINE(qdev_get_machine());
709     const char *boot_device = machine->boot_order;
710     int ret, i;
711     size_t cb = 0;
712     char *bootlist;
713     void *fdt;
714     sPAPRPHBState *phb;
715 
716     fdt = g_malloc(FDT_MAX_SIZE);
717 
718     /* open out the base tree into a temp buffer for the final tweaks */
719     _FDT((fdt_open_into(spapr->fdt_skel, fdt, FDT_MAX_SIZE)));
720 
721     ret = spapr_populate_memory(spapr, fdt);
722     if (ret < 0) {
723         fprintf(stderr, "couldn't setup memory nodes in fdt\n");
724         exit(1);
725     }
726 
727     ret = spapr_populate_vdevice(spapr->vio_bus, fdt);
728     if (ret < 0) {
729         fprintf(stderr, "couldn't setup vio devices in fdt\n");
730         exit(1);
731     }
732 
733     QLIST_FOREACH(phb, &spapr->phbs, list) {
734         ret = spapr_populate_pci_dt(phb, PHANDLE_XICP, fdt);
735     }
736 
737     if (ret < 0) {
738         fprintf(stderr, "couldn't setup PCI devices in fdt\n");
739         exit(1);
740     }
741 
742     /* RTAS */
743     ret = spapr_rtas_device_tree_setup(fdt, rtas_addr, rtas_size);
744     if (ret < 0) {
745         fprintf(stderr, "Couldn't set up RTAS device tree properties\n");
746     }
747 
748     /* Advertise NUMA via ibm,associativity */
749     ret = spapr_fixup_cpu_dt(fdt, spapr);
750     if (ret < 0) {
751         fprintf(stderr, "Couldn't finalize CPU device tree properties\n");
752     }
753 
754     bootlist = get_boot_devices_list(&cb, true);
755     if (cb && bootlist) {
756         int offset = fdt_path_offset(fdt, "/chosen");
757         if (offset < 0) {
758             exit(1);
759         }
760         for (i = 0; i < cb; i++) {
761             if (bootlist[i] == '\n') {
762                 bootlist[i] = ' ';
763             }
764 
765         }
766         ret = fdt_setprop_string(fdt, offset, "qemu,boot-list", bootlist);
767     }
768 
769     if (boot_device && strlen(boot_device)) {
770         int offset = fdt_path_offset(fdt, "/chosen");
771 
772         if (offset < 0) {
773             exit(1);
774         }
775         fdt_setprop_string(fdt, offset, "qemu,boot-device", boot_device);
776     }
777 
778     if (!spapr->has_graphics) {
779         spapr_populate_chosen_stdout(fdt, spapr->vio_bus);
780     }
781 
782     _FDT((fdt_pack(fdt)));
783 
784     if (fdt_totalsize(fdt) > FDT_MAX_SIZE) {
785         error_report("FDT too big ! 0x%x bytes (max is 0x%x)",
786                      fdt_totalsize(fdt), FDT_MAX_SIZE);
787         exit(1);
788     }
789 
790     cpu_physical_memory_write(fdt_addr, fdt, fdt_totalsize(fdt));
791 
792     g_free(bootlist);
793     g_free(fdt);
794 }
795 
796 static uint64_t translate_kernel_address(void *opaque, uint64_t addr)
797 {
798     return (addr & 0x0fffffff) + KERNEL_LOAD_ADDR;
799 }
800 
801 static void emulate_spapr_hypercall(PowerPCCPU *cpu)
802 {
803     CPUPPCState *env = &cpu->env;
804 
805     if (msr_pr) {
806         hcall_dprintf("Hypercall made with MSR[PR]=1\n");
807         env->gpr[3] = H_PRIVILEGE;
808     } else {
809         env->gpr[3] = spapr_hypercall(cpu, env->gpr[3], &env->gpr[4]);
810     }
811 }
812 
813 #define HPTE(_table, _i)   (void *)(((uint64_t *)(_table)) + ((_i) * 2))
814 #define HPTE_VALID(_hpte)  (tswap64(*((uint64_t *)(_hpte))) & HPTE64_V_VALID)
815 #define HPTE_DIRTY(_hpte)  (tswap64(*((uint64_t *)(_hpte))) & HPTE64_V_HPTE_DIRTY)
816 #define CLEAN_HPTE(_hpte)  ((*(uint64_t *)(_hpte)) &= tswap64(~HPTE64_V_HPTE_DIRTY))
817 #define DIRTY_HPTE(_hpte)  ((*(uint64_t *)(_hpte)) |= tswap64(HPTE64_V_HPTE_DIRTY))
818 
819 static void spapr_reset_htab(sPAPRMachineState *spapr)
820 {
821     long shift;
822     int index;
823 
824     /* allocate hash page table.  For now we always make this 16mb,
825      * later we should probably make it scale to the size of guest
826      * RAM */
827 
828     shift = kvmppc_reset_htab(spapr->htab_shift);
829 
830     if (shift > 0) {
831         /* Kernel handles htab, we don't need to allocate one */
832         spapr->htab_shift = shift;
833         kvmppc_kern_htab = true;
834 
835         /* Tell readers to update their file descriptor */
836         if (spapr->htab_fd >= 0) {
837             spapr->htab_fd_stale = true;
838         }
839     } else {
840         if (!spapr->htab) {
841             /* Allocate an htab if we don't yet have one */
842             spapr->htab = qemu_memalign(HTAB_SIZE(spapr), HTAB_SIZE(spapr));
843         }
844 
845         /* And clear it */
846         memset(spapr->htab, 0, HTAB_SIZE(spapr));
847 
848         for (index = 0; index < HTAB_SIZE(spapr) / HASH_PTE_SIZE_64; index++) {
849             DIRTY_HPTE(HPTE(spapr->htab, index));
850         }
851     }
852 
853     /* Update the RMA size if necessary */
854     if (spapr->vrma_adjust) {
855         spapr->rma_size = kvmppc_rma_size(spapr_node0_size(),
856                                           spapr->htab_shift);
857     }
858 }
859 
860 static int find_unknown_sysbus_device(SysBusDevice *sbdev, void *opaque)
861 {
862     bool matched = false;
863 
864     if (object_dynamic_cast(OBJECT(sbdev), TYPE_SPAPR_PCI_HOST_BRIDGE)) {
865         matched = true;
866     }
867 
868     if (!matched) {
869         error_report("Device %s is not supported by this machine yet.",
870                      qdev_fw_name(DEVICE(sbdev)));
871         exit(1);
872     }
873 
874     return 0;
875 }
876 
877 /*
878  * A guest reset will cause spapr->htab_fd to become stale if being used.
879  * Reopen the file descriptor to make sure the whole HTAB is properly read.
880  */
881 static int spapr_check_htab_fd(sPAPRMachineState *spapr)
882 {
883     int rc = 0;
884 
885     if (spapr->htab_fd_stale) {
886         close(spapr->htab_fd);
887         spapr->htab_fd = kvmppc_get_htab_fd(false);
888         if (spapr->htab_fd < 0) {
889             error_report("Unable to open fd for reading hash table from KVM: "
890                          "%s", strerror(errno));
891             rc = -1;
892         }
893         spapr->htab_fd_stale = false;
894     }
895 
896     return rc;
897 }
898 
899 static void ppc_spapr_reset(void)
900 {
901     sPAPRMachineState *spapr = SPAPR_MACHINE(qdev_get_machine());
902     PowerPCCPU *first_ppc_cpu;
903     uint32_t rtas_limit;
904 
905     /* Check for unknown sysbus devices */
906     foreach_dynamic_sysbus_device(find_unknown_sysbus_device, NULL);
907 
908     /* Reset the hash table & recalc the RMA */
909     spapr_reset_htab(spapr);
910 
911     qemu_devices_reset();
912 
913     /*
914      * We place the device tree and RTAS just below either the top of the RMA,
915      * or just below 2GB, whichever is lowere, so that it can be
916      * processed with 32-bit real mode code if necessary
917      */
918     rtas_limit = MIN(spapr->rma_size, RTAS_MAX_ADDR);
919     spapr->rtas_addr = rtas_limit - RTAS_MAX_SIZE;
920     spapr->fdt_addr = spapr->rtas_addr - FDT_MAX_SIZE;
921 
922     /* Load the fdt */
923     spapr_finalize_fdt(spapr, spapr->fdt_addr, spapr->rtas_addr,
924                        spapr->rtas_size);
925 
926     /* Copy RTAS over */
927     cpu_physical_memory_write(spapr->rtas_addr, spapr->rtas_blob,
928                               spapr->rtas_size);
929 
930     /* Set up the entry state */
931     first_ppc_cpu = POWERPC_CPU(first_cpu);
932     first_ppc_cpu->env.gpr[3] = spapr->fdt_addr;
933     first_ppc_cpu->env.gpr[5] = 0;
934     first_cpu->halted = 0;
935     first_ppc_cpu->env.nip = SPAPR_ENTRY_POINT;
936 
937 }
938 
939 static void spapr_cpu_reset(void *opaque)
940 {
941     sPAPRMachineState *spapr = SPAPR_MACHINE(qdev_get_machine());
942     PowerPCCPU *cpu = opaque;
943     CPUState *cs = CPU(cpu);
944     CPUPPCState *env = &cpu->env;
945 
946     cpu_reset(cs);
947 
948     /* All CPUs start halted.  CPU0 is unhalted from the machine level
949      * reset code and the rest are explicitly started up by the guest
950      * using an RTAS call */
951     cs->halted = 1;
952 
953     env->spr[SPR_HIOR] = 0;
954 
955     env->external_htab = (uint8_t *)spapr->htab;
956     if (kvm_enabled() && !env->external_htab) {
957         /*
958          * HV KVM, set external_htab to 1 so our ppc_hash64_load_hpte*
959          * functions do the right thing.
960          */
961         env->external_htab = (void *)1;
962     }
963     env->htab_base = -1;
964     /*
965      * htab_mask is the mask used to normalize hash value to PTEG index.
966      * htab_shift is log2 of hash table size.
967      * We have 8 hpte per group, and each hpte is 16 bytes.
968      * ie have 128 bytes per hpte entry.
969      */
970     env->htab_mask = (1ULL << (spapr->htab_shift - 7)) - 1;
971     env->spr[SPR_SDR1] = (target_ulong)(uintptr_t)spapr->htab |
972         (spapr->htab_shift - 18);
973 }
974 
975 static void spapr_create_nvram(sPAPRMachineState *spapr)
976 {
977     DeviceState *dev = qdev_create(&spapr->vio_bus->bus, "spapr-nvram");
978     DriveInfo *dinfo = drive_get(IF_PFLASH, 0, 0);
979 
980     if (dinfo) {
981         qdev_prop_set_drive_nofail(dev, "drive", blk_by_legacy_dinfo(dinfo));
982     }
983 
984     qdev_init_nofail(dev);
985 
986     spapr->nvram = (struct sPAPRNVRAM *)dev;
987 }
988 
989 static void spapr_rtc_create(sPAPRMachineState *spapr)
990 {
991     DeviceState *dev = qdev_create(NULL, TYPE_SPAPR_RTC);
992 
993     qdev_init_nofail(dev);
994     spapr->rtc = dev;
995 
996     object_property_add_alias(qdev_get_machine(), "rtc-time",
997                               OBJECT(spapr->rtc), "date", NULL);
998 }
999 
1000 /* Returns whether we want to use VGA or not */
1001 static int spapr_vga_init(PCIBus *pci_bus)
1002 {
1003     switch (vga_interface_type) {
1004     case VGA_NONE:
1005         return false;
1006     case VGA_DEVICE:
1007         return true;
1008     case VGA_STD:
1009         return pci_vga_init(pci_bus) != NULL;
1010     default:
1011         fprintf(stderr, "This vga model is not supported,"
1012                 "currently it only supports -vga std\n");
1013         exit(0);
1014     }
1015 }
1016 
1017 static int spapr_post_load(void *opaque, int version_id)
1018 {
1019     sPAPRMachineState *spapr = (sPAPRMachineState *)opaque;
1020     int err = 0;
1021 
1022     /* In earlier versions, there was no separate qdev for the PAPR
1023      * RTC, so the RTC offset was stored directly in sPAPREnvironment.
1024      * So when migrating from those versions, poke the incoming offset
1025      * value into the RTC device */
1026     if (version_id < 3) {
1027         err = spapr_rtc_import_offset(spapr->rtc, spapr->rtc_offset);
1028     }
1029 
1030     return err;
1031 }
1032 
1033 static bool version_before_3(void *opaque, int version_id)
1034 {
1035     return version_id < 3;
1036 }
1037 
1038 static const VMStateDescription vmstate_spapr = {
1039     .name = "spapr",
1040     .version_id = 3,
1041     .minimum_version_id = 1,
1042     .post_load = spapr_post_load,
1043     .fields = (VMStateField[]) {
1044         /* used to be @next_irq */
1045         VMSTATE_UNUSED_BUFFER(version_before_3, 0, 4),
1046 
1047         /* RTC offset */
1048         VMSTATE_UINT64_TEST(rtc_offset, sPAPRMachineState, version_before_3),
1049 
1050         VMSTATE_PPC_TIMEBASE_V(tb, sPAPRMachineState, 2),
1051         VMSTATE_END_OF_LIST()
1052     },
1053 };
1054 
1055 static int htab_save_setup(QEMUFile *f, void *opaque)
1056 {
1057     sPAPRMachineState *spapr = opaque;
1058 
1059     /* "Iteration" header */
1060     qemu_put_be32(f, spapr->htab_shift);
1061 
1062     if (spapr->htab) {
1063         spapr->htab_save_index = 0;
1064         spapr->htab_first_pass = true;
1065     } else {
1066         assert(kvm_enabled());
1067 
1068         spapr->htab_fd = kvmppc_get_htab_fd(false);
1069         spapr->htab_fd_stale = false;
1070         if (spapr->htab_fd < 0) {
1071             fprintf(stderr, "Unable to open fd for reading hash table from KVM: %s\n",
1072                     strerror(errno));
1073             return -1;
1074         }
1075     }
1076 
1077 
1078     return 0;
1079 }
1080 
1081 static void htab_save_first_pass(QEMUFile *f, sPAPRMachineState *spapr,
1082                                  int64_t max_ns)
1083 {
1084     int htabslots = HTAB_SIZE(spapr) / HASH_PTE_SIZE_64;
1085     int index = spapr->htab_save_index;
1086     int64_t starttime = qemu_clock_get_ns(QEMU_CLOCK_REALTIME);
1087 
1088     assert(spapr->htab_first_pass);
1089 
1090     do {
1091         int chunkstart;
1092 
1093         /* Consume invalid HPTEs */
1094         while ((index < htabslots)
1095                && !HPTE_VALID(HPTE(spapr->htab, index))) {
1096             index++;
1097             CLEAN_HPTE(HPTE(spapr->htab, index));
1098         }
1099 
1100         /* Consume valid HPTEs */
1101         chunkstart = index;
1102         while ((index < htabslots) && (index - chunkstart < USHRT_MAX)
1103                && HPTE_VALID(HPTE(spapr->htab, index))) {
1104             index++;
1105             CLEAN_HPTE(HPTE(spapr->htab, index));
1106         }
1107 
1108         if (index > chunkstart) {
1109             int n_valid = index - chunkstart;
1110 
1111             qemu_put_be32(f, chunkstart);
1112             qemu_put_be16(f, n_valid);
1113             qemu_put_be16(f, 0);
1114             qemu_put_buffer(f, HPTE(spapr->htab, chunkstart),
1115                             HASH_PTE_SIZE_64 * n_valid);
1116 
1117             if ((qemu_clock_get_ns(QEMU_CLOCK_REALTIME) - starttime) > max_ns) {
1118                 break;
1119             }
1120         }
1121     } while ((index < htabslots) && !qemu_file_rate_limit(f));
1122 
1123     if (index >= htabslots) {
1124         assert(index == htabslots);
1125         index = 0;
1126         spapr->htab_first_pass = false;
1127     }
1128     spapr->htab_save_index = index;
1129 }
1130 
1131 static int htab_save_later_pass(QEMUFile *f, sPAPRMachineState *spapr,
1132                                 int64_t max_ns)
1133 {
1134     bool final = max_ns < 0;
1135     int htabslots = HTAB_SIZE(spapr) / HASH_PTE_SIZE_64;
1136     int examined = 0, sent = 0;
1137     int index = spapr->htab_save_index;
1138     int64_t starttime = qemu_clock_get_ns(QEMU_CLOCK_REALTIME);
1139 
1140     assert(!spapr->htab_first_pass);
1141 
1142     do {
1143         int chunkstart, invalidstart;
1144 
1145         /* Consume non-dirty HPTEs */
1146         while ((index < htabslots)
1147                && !HPTE_DIRTY(HPTE(spapr->htab, index))) {
1148             index++;
1149             examined++;
1150         }
1151 
1152         chunkstart = index;
1153         /* Consume valid dirty HPTEs */
1154         while ((index < htabslots) && (index - chunkstart < USHRT_MAX)
1155                && HPTE_DIRTY(HPTE(spapr->htab, index))
1156                && HPTE_VALID(HPTE(spapr->htab, index))) {
1157             CLEAN_HPTE(HPTE(spapr->htab, index));
1158             index++;
1159             examined++;
1160         }
1161 
1162         invalidstart = index;
1163         /* Consume invalid dirty HPTEs */
1164         while ((index < htabslots) && (index - invalidstart < USHRT_MAX)
1165                && HPTE_DIRTY(HPTE(spapr->htab, index))
1166                && !HPTE_VALID(HPTE(spapr->htab, index))) {
1167             CLEAN_HPTE(HPTE(spapr->htab, index));
1168             index++;
1169             examined++;
1170         }
1171 
1172         if (index > chunkstart) {
1173             int n_valid = invalidstart - chunkstart;
1174             int n_invalid = index - invalidstart;
1175 
1176             qemu_put_be32(f, chunkstart);
1177             qemu_put_be16(f, n_valid);
1178             qemu_put_be16(f, n_invalid);
1179             qemu_put_buffer(f, HPTE(spapr->htab, chunkstart),
1180                             HASH_PTE_SIZE_64 * n_valid);
1181             sent += index - chunkstart;
1182 
1183             if (!final && (qemu_clock_get_ns(QEMU_CLOCK_REALTIME) - starttime) > max_ns) {
1184                 break;
1185             }
1186         }
1187 
1188         if (examined >= htabslots) {
1189             break;
1190         }
1191 
1192         if (index >= htabslots) {
1193             assert(index == htabslots);
1194             index = 0;
1195         }
1196     } while ((examined < htabslots) && (!qemu_file_rate_limit(f) || final));
1197 
1198     if (index >= htabslots) {
1199         assert(index == htabslots);
1200         index = 0;
1201     }
1202 
1203     spapr->htab_save_index = index;
1204 
1205     return (examined >= htabslots) && (sent == 0) ? 1 : 0;
1206 }
1207 
1208 #define MAX_ITERATION_NS    5000000 /* 5 ms */
1209 #define MAX_KVM_BUF_SIZE    2048
1210 
1211 static int htab_save_iterate(QEMUFile *f, void *opaque)
1212 {
1213     sPAPRMachineState *spapr = opaque;
1214     int rc = 0;
1215 
1216     /* Iteration header */
1217     qemu_put_be32(f, 0);
1218 
1219     if (!spapr->htab) {
1220         assert(kvm_enabled());
1221 
1222         rc = spapr_check_htab_fd(spapr);
1223         if (rc < 0) {
1224             return rc;
1225         }
1226 
1227         rc = kvmppc_save_htab(f, spapr->htab_fd,
1228                               MAX_KVM_BUF_SIZE, MAX_ITERATION_NS);
1229         if (rc < 0) {
1230             return rc;
1231         }
1232     } else  if (spapr->htab_first_pass) {
1233         htab_save_first_pass(f, spapr, MAX_ITERATION_NS);
1234     } else {
1235         rc = htab_save_later_pass(f, spapr, MAX_ITERATION_NS);
1236     }
1237 
1238     /* End marker */
1239     qemu_put_be32(f, 0);
1240     qemu_put_be16(f, 0);
1241     qemu_put_be16(f, 0);
1242 
1243     return rc;
1244 }
1245 
1246 static int htab_save_complete(QEMUFile *f, void *opaque)
1247 {
1248     sPAPRMachineState *spapr = opaque;
1249 
1250     /* Iteration header */
1251     qemu_put_be32(f, 0);
1252 
1253     if (!spapr->htab) {
1254         int rc;
1255 
1256         assert(kvm_enabled());
1257 
1258         rc = spapr_check_htab_fd(spapr);
1259         if (rc < 0) {
1260             return rc;
1261         }
1262 
1263         rc = kvmppc_save_htab(f, spapr->htab_fd, MAX_KVM_BUF_SIZE, -1);
1264         if (rc < 0) {
1265             return rc;
1266         }
1267         close(spapr->htab_fd);
1268         spapr->htab_fd = -1;
1269     } else {
1270         htab_save_later_pass(f, spapr, -1);
1271     }
1272 
1273     /* End marker */
1274     qemu_put_be32(f, 0);
1275     qemu_put_be16(f, 0);
1276     qemu_put_be16(f, 0);
1277 
1278     return 0;
1279 }
1280 
1281 static int htab_load(QEMUFile *f, void *opaque, int version_id)
1282 {
1283     sPAPRMachineState *spapr = opaque;
1284     uint32_t section_hdr;
1285     int fd = -1;
1286 
1287     if (version_id < 1 || version_id > 1) {
1288         fprintf(stderr, "htab_load() bad version\n");
1289         return -EINVAL;
1290     }
1291 
1292     section_hdr = qemu_get_be32(f);
1293 
1294     if (section_hdr) {
1295         /* First section, just the hash shift */
1296         if (spapr->htab_shift != section_hdr) {
1297             return -EINVAL;
1298         }
1299         return 0;
1300     }
1301 
1302     if (!spapr->htab) {
1303         assert(kvm_enabled());
1304 
1305         fd = kvmppc_get_htab_fd(true);
1306         if (fd < 0) {
1307             fprintf(stderr, "Unable to open fd to restore KVM hash table: %s\n",
1308                     strerror(errno));
1309         }
1310     }
1311 
1312     while (true) {
1313         uint32_t index;
1314         uint16_t n_valid, n_invalid;
1315 
1316         index = qemu_get_be32(f);
1317         n_valid = qemu_get_be16(f);
1318         n_invalid = qemu_get_be16(f);
1319 
1320         if ((index == 0) && (n_valid == 0) && (n_invalid == 0)) {
1321             /* End of Stream */
1322             break;
1323         }
1324 
1325         if ((index + n_valid + n_invalid) >
1326             (HTAB_SIZE(spapr) / HASH_PTE_SIZE_64)) {
1327             /* Bad index in stream */
1328             fprintf(stderr, "htab_load() bad index %d (%hd+%hd entries) "
1329                     "in htab stream (htab_shift=%d)\n", index, n_valid, n_invalid,
1330                     spapr->htab_shift);
1331             return -EINVAL;
1332         }
1333 
1334         if (spapr->htab) {
1335             if (n_valid) {
1336                 qemu_get_buffer(f, HPTE(spapr->htab, index),
1337                                 HASH_PTE_SIZE_64 * n_valid);
1338             }
1339             if (n_invalid) {
1340                 memset(HPTE(spapr->htab, index + n_valid), 0,
1341                        HASH_PTE_SIZE_64 * n_invalid);
1342             }
1343         } else {
1344             int rc;
1345 
1346             assert(fd >= 0);
1347 
1348             rc = kvmppc_load_htab_chunk(f, fd, index, n_valid, n_invalid);
1349             if (rc < 0) {
1350                 return rc;
1351             }
1352         }
1353     }
1354 
1355     if (!spapr->htab) {
1356         assert(fd >= 0);
1357         close(fd);
1358     }
1359 
1360     return 0;
1361 }
1362 
1363 static SaveVMHandlers savevm_htab_handlers = {
1364     .save_live_setup = htab_save_setup,
1365     .save_live_iterate = htab_save_iterate,
1366     .save_live_complete = htab_save_complete,
1367     .load_state = htab_load,
1368 };
1369 
1370 static void spapr_boot_set(void *opaque, const char *boot_device,
1371                            Error **errp)
1372 {
1373     MachineState *machine = MACHINE(qdev_get_machine());
1374     machine->boot_order = g_strdup(boot_device);
1375 }
1376 
1377 /* pSeries LPAR / sPAPR hardware init */
1378 static void ppc_spapr_init(MachineState *machine)
1379 {
1380     sPAPRMachineState *spapr = SPAPR_MACHINE(machine);
1381     const char *cpu_model = machine->cpu_model;
1382     const char *kernel_filename = machine->kernel_filename;
1383     const char *kernel_cmdline = machine->kernel_cmdline;
1384     const char *initrd_filename = machine->initrd_filename;
1385     PowerPCCPU *cpu;
1386     CPUPPCState *env;
1387     PCIHostState *phb;
1388     int i;
1389     MemoryRegion *sysmem = get_system_memory();
1390     MemoryRegion *ram = g_new(MemoryRegion, 1);
1391     MemoryRegion *rma_region;
1392     void *rma = NULL;
1393     hwaddr rma_alloc_size;
1394     hwaddr node0_size = spapr_node0_size();
1395     uint32_t initrd_base = 0;
1396     long kernel_size = 0, initrd_size = 0;
1397     long load_limit, fw_size;
1398     bool kernel_le = false;
1399     char *filename;
1400 
1401     msi_supported = true;
1402 
1403     QLIST_INIT(&spapr->phbs);
1404 
1405     cpu_ppc_hypercall = emulate_spapr_hypercall;
1406 
1407     /* Allocate RMA if necessary */
1408     rma_alloc_size = kvmppc_alloc_rma(&rma);
1409 
1410     if (rma_alloc_size == -1) {
1411         error_report("Unable to create RMA");
1412         exit(1);
1413     }
1414 
1415     if (rma_alloc_size && (rma_alloc_size < node0_size)) {
1416         spapr->rma_size = rma_alloc_size;
1417     } else {
1418         spapr->rma_size = node0_size;
1419 
1420         /* With KVM, we don't actually know whether KVM supports an
1421          * unbounded RMA (PR KVM) or is limited by the hash table size
1422          * (HV KVM using VRMA), so we always assume the latter
1423          *
1424          * In that case, we also limit the initial allocations for RTAS
1425          * etc... to 256M since we have no way to know what the VRMA size
1426          * is going to be as it depends on the size of the hash table
1427          * isn't determined yet.
1428          */
1429         if (kvm_enabled()) {
1430             spapr->vrma_adjust = 1;
1431             spapr->rma_size = MIN(spapr->rma_size, 0x10000000);
1432         }
1433     }
1434 
1435     if (spapr->rma_size > node0_size) {
1436         fprintf(stderr, "Error: Numa node 0 has to span the RMA (%#08"HWADDR_PRIx")\n",
1437                 spapr->rma_size);
1438         exit(1);
1439     }
1440 
1441     /* Setup a load limit for the ramdisk leaving room for SLOF and FDT */
1442     load_limit = MIN(spapr->rma_size, RTAS_MAX_ADDR) - FW_OVERHEAD;
1443 
1444     /* We aim for a hash table of size 1/128 the size of RAM.  The
1445      * normal rule of thumb is 1/64 the size of RAM, but that's much
1446      * more than needed for the Linux guests we support. */
1447     spapr->htab_shift = 18; /* Minimum architected size */
1448     while (spapr->htab_shift <= 46) {
1449         if ((1ULL << (spapr->htab_shift + 7)) >= machine->ram_size) {
1450             break;
1451         }
1452         spapr->htab_shift++;
1453     }
1454 
1455     /* Set up Interrupt Controller before we create the VCPUs */
1456     spapr->icp = xics_system_init(machine,
1457                                   DIV_ROUND_UP(max_cpus * kvmppc_smt_threads(),
1458                                                smp_threads),
1459                                   XICS_IRQS);
1460 
1461     /* init CPUs */
1462     if (cpu_model == NULL) {
1463         cpu_model = kvm_enabled() ? "host" : "POWER7";
1464     }
1465     for (i = 0; i < smp_cpus; i++) {
1466         cpu = cpu_ppc_init(cpu_model);
1467         if (cpu == NULL) {
1468             fprintf(stderr, "Unable to find PowerPC CPU definition\n");
1469             exit(1);
1470         }
1471         env = &cpu->env;
1472 
1473         /* Set time-base frequency to 512 MHz */
1474         cpu_ppc_tb_init(env, TIMEBASE_FREQ);
1475 
1476         /* PAPR always has exception vectors in RAM not ROM. To ensure this,
1477          * MSR[IP] should never be set.
1478          */
1479         env->msr_mask &= ~(1 << 6);
1480 
1481         /* Tell KVM that we're in PAPR mode */
1482         if (kvm_enabled()) {
1483             kvmppc_set_papr(cpu);
1484         }
1485 
1486         if (cpu->max_compat) {
1487             if (ppc_set_compat(cpu, cpu->max_compat) < 0) {
1488                 exit(1);
1489             }
1490         }
1491 
1492         xics_cpu_setup(spapr->icp, cpu);
1493 
1494         qemu_register_reset(spapr_cpu_reset, cpu);
1495     }
1496 
1497     if (kvm_enabled()) {
1498         /* Enable H_LOGICAL_CI_* so SLOF can talk to in-kernel devices */
1499         kvmppc_enable_logical_ci_hcalls();
1500     }
1501 
1502     /* allocate RAM */
1503     memory_region_allocate_system_memory(ram, NULL, "ppc_spapr.ram",
1504                                          machine->ram_size);
1505     memory_region_add_subregion(sysmem, 0, ram);
1506 
1507     if (rma_alloc_size && rma) {
1508         rma_region = g_new(MemoryRegion, 1);
1509         memory_region_init_ram_ptr(rma_region, NULL, "ppc_spapr.rma",
1510                                    rma_alloc_size, rma);
1511         vmstate_register_ram_global(rma_region);
1512         memory_region_add_subregion(sysmem, 0, rma_region);
1513     }
1514 
1515     filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, "spapr-rtas.bin");
1516     if (!filename) {
1517         error_report("Could not find LPAR rtas '%s'", "spapr-rtas.bin");
1518         exit(1);
1519     }
1520     spapr->rtas_size = get_image_size(filename);
1521     spapr->rtas_blob = g_malloc(spapr->rtas_size);
1522     if (load_image_size(filename, spapr->rtas_blob, spapr->rtas_size) < 0) {
1523         error_report("Could not load LPAR rtas '%s'", filename);
1524         exit(1);
1525     }
1526     if (spapr->rtas_size > RTAS_MAX_SIZE) {
1527         error_report("RTAS too big ! 0x%zx bytes (max is 0x%x)",
1528                      (size_t)spapr->rtas_size, RTAS_MAX_SIZE);
1529         exit(1);
1530     }
1531     g_free(filename);
1532 
1533     /* Set up EPOW events infrastructure */
1534     spapr_events_init(spapr);
1535 
1536     /* Set up the RTC RTAS interfaces */
1537     spapr_rtc_create(spapr);
1538 
1539     /* Set up VIO bus */
1540     spapr->vio_bus = spapr_vio_bus_init();
1541 
1542     for (i = 0; i < MAX_SERIAL_PORTS; i++) {
1543         if (serial_hds[i]) {
1544             spapr_vty_create(spapr->vio_bus, serial_hds[i]);
1545         }
1546     }
1547 
1548     /* We always have at least the nvram device on VIO */
1549     spapr_create_nvram(spapr);
1550 
1551     /* Set up PCI */
1552     spapr_pci_rtas_init();
1553 
1554     phb = spapr_create_phb(spapr, 0);
1555 
1556     for (i = 0; i < nb_nics; i++) {
1557         NICInfo *nd = &nd_table[i];
1558 
1559         if (!nd->model) {
1560             nd->model = g_strdup("ibmveth");
1561         }
1562 
1563         if (strcmp(nd->model, "ibmveth") == 0) {
1564             spapr_vlan_create(spapr->vio_bus, nd);
1565         } else {
1566             pci_nic_init_nofail(&nd_table[i], phb->bus, nd->model, NULL);
1567         }
1568     }
1569 
1570     for (i = 0; i <= drive_get_max_bus(IF_SCSI); i++) {
1571         spapr_vscsi_create(spapr->vio_bus);
1572     }
1573 
1574     /* Graphics */
1575     if (spapr_vga_init(phb->bus)) {
1576         spapr->has_graphics = true;
1577         machine->usb |= defaults_enabled() && !machine->usb_disabled;
1578     }
1579 
1580     if (machine->usb) {
1581         pci_create_simple(phb->bus, -1, "pci-ohci");
1582 
1583         if (spapr->has_graphics) {
1584             USBBus *usb_bus = usb_bus_find(-1);
1585 
1586             usb_create_simple(usb_bus, "usb-kbd");
1587             usb_create_simple(usb_bus, "usb-mouse");
1588         }
1589     }
1590 
1591     if (spapr->rma_size < (MIN_RMA_SLOF << 20)) {
1592         fprintf(stderr, "qemu: pSeries SLOF firmware requires >= "
1593                 "%ldM guest RMA (Real Mode Area memory)\n", MIN_RMA_SLOF);
1594         exit(1);
1595     }
1596 
1597     if (kernel_filename) {
1598         uint64_t lowaddr = 0;
1599 
1600         kernel_size = load_elf(kernel_filename, translate_kernel_address, NULL,
1601                                NULL, &lowaddr, NULL, 1, ELF_MACHINE, 0);
1602         if (kernel_size == ELF_LOAD_WRONG_ENDIAN) {
1603             kernel_size = load_elf(kernel_filename,
1604                                    translate_kernel_address, NULL,
1605                                    NULL, &lowaddr, NULL, 0, ELF_MACHINE, 0);
1606             kernel_le = kernel_size > 0;
1607         }
1608         if (kernel_size < 0) {
1609             fprintf(stderr, "qemu: error loading %s: %s\n",
1610                     kernel_filename, load_elf_strerror(kernel_size));
1611             exit(1);
1612         }
1613 
1614         /* load initrd */
1615         if (initrd_filename) {
1616             /* Try to locate the initrd in the gap between the kernel
1617              * and the firmware. Add a bit of space just in case
1618              */
1619             initrd_base = (KERNEL_LOAD_ADDR + kernel_size + 0x1ffff) & ~0xffff;
1620             initrd_size = load_image_targphys(initrd_filename, initrd_base,
1621                                               load_limit - initrd_base);
1622             if (initrd_size < 0) {
1623                 fprintf(stderr, "qemu: could not load initial ram disk '%s'\n",
1624                         initrd_filename);
1625                 exit(1);
1626             }
1627         } else {
1628             initrd_base = 0;
1629             initrd_size = 0;
1630         }
1631     }
1632 
1633     if (bios_name == NULL) {
1634         bios_name = FW_FILE_NAME;
1635     }
1636     filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, bios_name);
1637     if (!filename) {
1638         error_report("Could not find LPAR firmware '%s'", bios_name);
1639         exit(1);
1640     }
1641     fw_size = load_image_targphys(filename, 0, FW_MAX_SIZE);
1642     if (fw_size <= 0) {
1643         error_report("Could not load LPAR firmware '%s'", filename);
1644         exit(1);
1645     }
1646     g_free(filename);
1647 
1648     /* FIXME: Should register things through the MachineState's qdev
1649      * interface, this is a legacy from the sPAPREnvironment structure
1650      * which predated MachineState but had a similar function */
1651     vmstate_register(NULL, 0, &vmstate_spapr, spapr);
1652     register_savevm_live(NULL, "spapr/htab", -1, 1,
1653                          &savevm_htab_handlers, spapr);
1654 
1655     /* Prepare the device tree */
1656     spapr->fdt_skel = spapr_create_fdt_skel(initrd_base, initrd_size,
1657                                             kernel_size, kernel_le,
1658                                             kernel_cmdline,
1659                                             spapr->check_exception_irq);
1660     assert(spapr->fdt_skel != NULL);
1661 
1662     /* used by RTAS */
1663     QTAILQ_INIT(&spapr->ccs_list);
1664     qemu_register_reset(spapr_ccs_reset_hook, spapr);
1665 
1666     qemu_register_boot_set(spapr_boot_set, spapr);
1667 }
1668 
1669 static int spapr_kvm_type(const char *vm_type)
1670 {
1671     if (!vm_type) {
1672         return 0;
1673     }
1674 
1675     if (!strcmp(vm_type, "HV")) {
1676         return 1;
1677     }
1678 
1679     if (!strcmp(vm_type, "PR")) {
1680         return 2;
1681     }
1682 
1683     error_report("Unknown kvm-type specified '%s'", vm_type);
1684     exit(1);
1685 }
1686 
1687 /*
1688  * Implementation of an interface to adjust firmware path
1689  * for the bootindex property handling.
1690  */
1691 static char *spapr_get_fw_dev_path(FWPathProvider *p, BusState *bus,
1692                                    DeviceState *dev)
1693 {
1694 #define CAST(type, obj, name) \
1695     ((type *)object_dynamic_cast(OBJECT(obj), (name)))
1696     SCSIDevice *d = CAST(SCSIDevice,  dev, TYPE_SCSI_DEVICE);
1697     sPAPRPHBState *phb = CAST(sPAPRPHBState, dev, TYPE_SPAPR_PCI_HOST_BRIDGE);
1698 
1699     if (d) {
1700         void *spapr = CAST(void, bus->parent, "spapr-vscsi");
1701         VirtIOSCSI *virtio = CAST(VirtIOSCSI, bus->parent, TYPE_VIRTIO_SCSI);
1702         USBDevice *usb = CAST(USBDevice, bus->parent, TYPE_USB_DEVICE);
1703 
1704         if (spapr) {
1705             /*
1706              * Replace "channel@0/disk@0,0" with "disk@8000000000000000":
1707              * We use SRP luns of the form 8000 | (bus << 8) | (id << 5) | lun
1708              * in the top 16 bits of the 64-bit LUN
1709              */
1710             unsigned id = 0x8000 | (d->id << 8) | d->lun;
1711             return g_strdup_printf("%s@%"PRIX64, qdev_fw_name(dev),
1712                                    (uint64_t)id << 48);
1713         } else if (virtio) {
1714             /*
1715              * We use SRP luns of the form 01000000 | (target << 8) | lun
1716              * in the top 32 bits of the 64-bit LUN
1717              * Note: the quote above is from SLOF and it is wrong,
1718              * the actual binding is:
1719              * swap 0100 or 10 << or 20 << ( target lun-id -- srplun )
1720              */
1721             unsigned id = 0x1000000 | (d->id << 16) | d->lun;
1722             return g_strdup_printf("%s@%"PRIX64, qdev_fw_name(dev),
1723                                    (uint64_t)id << 32);
1724         } else if (usb) {
1725             /*
1726              * We use SRP luns of the form 01000000 | (usb-port << 16) | lun
1727              * in the top 32 bits of the 64-bit LUN
1728              */
1729             unsigned usb_port = atoi(usb->port->path);
1730             unsigned id = 0x1000000 | (usb_port << 16) | d->lun;
1731             return g_strdup_printf("%s@%"PRIX64, qdev_fw_name(dev),
1732                                    (uint64_t)id << 32);
1733         }
1734     }
1735 
1736     if (phb) {
1737         /* Replace "pci" with "pci@800000020000000" */
1738         return g_strdup_printf("pci@%"PRIX64, phb->buid);
1739     }
1740 
1741     return NULL;
1742 }
1743 
1744 static char *spapr_get_kvm_type(Object *obj, Error **errp)
1745 {
1746     sPAPRMachineState *spapr = SPAPR_MACHINE(obj);
1747 
1748     return g_strdup(spapr->kvm_type);
1749 }
1750 
1751 static void spapr_set_kvm_type(Object *obj, const char *value, Error **errp)
1752 {
1753     sPAPRMachineState *spapr = SPAPR_MACHINE(obj);
1754 
1755     g_free(spapr->kvm_type);
1756     spapr->kvm_type = g_strdup(value);
1757 }
1758 
1759 static void spapr_machine_initfn(Object *obj)
1760 {
1761     object_property_add_str(obj, "kvm-type",
1762                             spapr_get_kvm_type, spapr_set_kvm_type, NULL);
1763     object_property_set_description(obj, "kvm-type",
1764                                     "Specifies the KVM virtualization mode (HV, PR)",
1765                                     NULL);
1766 }
1767 
1768 static void ppc_cpu_do_nmi_on_cpu(void *arg)
1769 {
1770     CPUState *cs = arg;
1771 
1772     cpu_synchronize_state(cs);
1773     ppc_cpu_do_system_reset(cs);
1774 }
1775 
1776 static void spapr_nmi(NMIState *n, int cpu_index, Error **errp)
1777 {
1778     CPUState *cs;
1779 
1780     CPU_FOREACH(cs) {
1781         async_run_on_cpu(cs, ppc_cpu_do_nmi_on_cpu, cs);
1782     }
1783 }
1784 
1785 static void spapr_machine_class_init(ObjectClass *oc, void *data)
1786 {
1787     MachineClass *mc = MACHINE_CLASS(oc);
1788     FWPathProviderClass *fwc = FW_PATH_PROVIDER_CLASS(oc);
1789     NMIClass *nc = NMI_CLASS(oc);
1790 
1791     mc->init = ppc_spapr_init;
1792     mc->reset = ppc_spapr_reset;
1793     mc->block_default_type = IF_SCSI;
1794     mc->max_cpus = MAX_CPUS;
1795     mc->no_parallel = 1;
1796     mc->default_boot_order = "";
1797     mc->default_ram_size = 512 * M_BYTE;
1798     mc->kvm_type = spapr_kvm_type;
1799     mc->has_dynamic_sysbus = true;
1800 
1801     fwc->get_dev_path = spapr_get_fw_dev_path;
1802     nc->nmi_monitor_handler = spapr_nmi;
1803 }
1804 
1805 static const TypeInfo spapr_machine_info = {
1806     .name          = TYPE_SPAPR_MACHINE,
1807     .parent        = TYPE_MACHINE,
1808     .abstract      = true,
1809     .instance_size = sizeof(sPAPRMachineState),
1810     .instance_init = spapr_machine_initfn,
1811     .class_size    = sizeof(sPAPRMachineClass),
1812     .class_init    = spapr_machine_class_init,
1813     .interfaces = (InterfaceInfo[]) {
1814         { TYPE_FW_PATH_PROVIDER },
1815         { TYPE_NMI },
1816         { }
1817     },
1818 };
1819 
1820 #define SPAPR_COMPAT_2_3 \
1821         HW_COMPAT_2_3 \
1822         {\
1823             .driver   = "spapr-pci-host-bridge",\
1824             .property = "dynamic-reconfiguration",\
1825             .value    = "off",\
1826         },
1827 
1828 #define SPAPR_COMPAT_2_2 \
1829         SPAPR_COMPAT_2_3 \
1830         HW_COMPAT_2_2 \
1831         {\
1832             .driver   = TYPE_SPAPR_PCI_HOST_BRIDGE,\
1833             .property = "mem_win_size",\
1834             .value    = "0x20000000",\
1835         },
1836 
1837 #define SPAPR_COMPAT_2_1 \
1838         SPAPR_COMPAT_2_2 \
1839         HW_COMPAT_2_1
1840 
1841 static void spapr_compat_2_3(Object *obj)
1842 {
1843 }
1844 
1845 static void spapr_compat_2_2(Object *obj)
1846 {
1847     spapr_compat_2_3(obj);
1848 }
1849 
1850 static void spapr_compat_2_1(Object *obj)
1851 {
1852     spapr_compat_2_2(obj);
1853 }
1854 
1855 static void spapr_machine_2_3_instance_init(Object *obj)
1856 {
1857     spapr_compat_2_3(obj);
1858     spapr_machine_initfn(obj);
1859 }
1860 
1861 static void spapr_machine_2_2_instance_init(Object *obj)
1862 {
1863     spapr_compat_2_2(obj);
1864     spapr_machine_initfn(obj);
1865 }
1866 
1867 static void spapr_machine_2_1_instance_init(Object *obj)
1868 {
1869     spapr_compat_2_1(obj);
1870     spapr_machine_initfn(obj);
1871 }
1872 
1873 static void spapr_machine_2_1_class_init(ObjectClass *oc, void *data)
1874 {
1875     MachineClass *mc = MACHINE_CLASS(oc);
1876     static GlobalProperty compat_props[] = {
1877         SPAPR_COMPAT_2_1
1878         { /* end of list */ }
1879     };
1880 
1881     mc->name = "pseries-2.1";
1882     mc->desc = "pSeries Logical Partition (PAPR compliant) v2.1";
1883     mc->compat_props = compat_props;
1884 }
1885 
1886 static const TypeInfo spapr_machine_2_1_info = {
1887     .name          = TYPE_SPAPR_MACHINE "2.1",
1888     .parent        = TYPE_SPAPR_MACHINE,
1889     .class_init    = spapr_machine_2_1_class_init,
1890     .instance_init = spapr_machine_2_1_instance_init,
1891 };
1892 
1893 static void spapr_machine_2_2_class_init(ObjectClass *oc, void *data)
1894 {
1895     static GlobalProperty compat_props[] = {
1896         SPAPR_COMPAT_2_2
1897         { /* end of list */ }
1898     };
1899     MachineClass *mc = MACHINE_CLASS(oc);
1900 
1901     mc->name = "pseries-2.2";
1902     mc->desc = "pSeries Logical Partition (PAPR compliant) v2.2";
1903     mc->compat_props = compat_props;
1904 }
1905 
1906 static const TypeInfo spapr_machine_2_2_info = {
1907     .name          = TYPE_SPAPR_MACHINE "2.2",
1908     .parent        = TYPE_SPAPR_MACHINE,
1909     .class_init    = spapr_machine_2_2_class_init,
1910     .instance_init = spapr_machine_2_2_instance_init,
1911 };
1912 
1913 static void spapr_machine_2_3_class_init(ObjectClass *oc, void *data)
1914 {
1915     static GlobalProperty compat_props[] = {
1916         SPAPR_COMPAT_2_3
1917         { /* end of list */ }
1918     };
1919     MachineClass *mc = MACHINE_CLASS(oc);
1920 
1921     mc->name = "pseries-2.3";
1922     mc->desc = "pSeries Logical Partition (PAPR compliant) v2.3";
1923     mc->compat_props = compat_props;
1924 }
1925 
1926 static const TypeInfo spapr_machine_2_3_info = {
1927     .name          = TYPE_SPAPR_MACHINE "2.3",
1928     .parent        = TYPE_SPAPR_MACHINE,
1929     .class_init    = spapr_machine_2_3_class_init,
1930     .instance_init = spapr_machine_2_3_instance_init,
1931 };
1932 
1933 static void spapr_machine_2_4_class_init(ObjectClass *oc, void *data)
1934 {
1935     MachineClass *mc = MACHINE_CLASS(oc);
1936 
1937     mc->name = "pseries-2.4";
1938     mc->desc = "pSeries Logical Partition (PAPR compliant) v2.4";
1939     mc->alias = "pseries";
1940     mc->is_default = 1;
1941 }
1942 
1943 static const TypeInfo spapr_machine_2_4_info = {
1944     .name          = TYPE_SPAPR_MACHINE "2.4",
1945     .parent        = TYPE_SPAPR_MACHINE,
1946     .class_init    = spapr_machine_2_4_class_init,
1947 };
1948 
1949 static void spapr_machine_register_types(void)
1950 {
1951     type_register_static(&spapr_machine_info);
1952     type_register_static(&spapr_machine_2_1_info);
1953     type_register_static(&spapr_machine_2_2_info);
1954     type_register_static(&spapr_machine_2_3_info);
1955     type_register_static(&spapr_machine_2_4_info);
1956 }
1957 
1958 type_init(spapr_machine_register_types)
1959