xref: /openbmc/qemu/hw/ppc/spapr.c (revision bcd82a96)
1 /*
2  * QEMU PowerPC pSeries Logical Partition (aka sPAPR) hardware System Emulator
3  *
4  * Copyright (c) 2004-2007 Fabrice Bellard
5  * Copyright (c) 2007 Jocelyn Mayer
6  * Copyright (c) 2010 David Gibson, IBM Corporation.
7  *
8  * Permission is hereby granted, free of charge, to any person obtaining a copy
9  * of this software and associated documentation files (the "Software"), to deal
10  * in the Software without restriction, including without limitation the rights
11  * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
12  * copies of the Software, and to permit persons to whom the Software is
13  * furnished to do so, subject to the following conditions:
14  *
15  * The above copyright notice and this permission notice shall be included in
16  * all copies or substantial portions of the Software.
17  *
18  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
19  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
20  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
21  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
22  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
23  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
24  * THE SOFTWARE.
25  *
26  */
27 #include "qemu/osdep.h"
28 #include "qapi/error.h"
29 #include "sysemu/sysemu.h"
30 #include "sysemu/numa.h"
31 #include "hw/hw.h"
32 #include "hw/fw-path-provider.h"
33 #include "elf.h"
34 #include "net/net.h"
35 #include "sysemu/device_tree.h"
36 #include "sysemu/block-backend.h"
37 #include "sysemu/cpus.h"
38 #include "sysemu/kvm.h"
39 #include "sysemu/device_tree.h"
40 #include "kvm_ppc.h"
41 #include "migration/migration.h"
42 #include "mmu-hash64.h"
43 #include "qom/cpu.h"
44 
45 #include "hw/boards.h"
46 #include "hw/ppc/ppc.h"
47 #include "hw/loader.h"
48 
49 #include "hw/ppc/spapr.h"
50 #include "hw/ppc/spapr_vio.h"
51 #include "hw/pci-host/spapr.h"
52 #include "hw/ppc/xics.h"
53 #include "hw/pci/msi.h"
54 
55 #include "hw/pci/pci.h"
56 #include "hw/scsi/scsi.h"
57 #include "hw/virtio/virtio-scsi.h"
58 
59 #include "exec/address-spaces.h"
60 #include "hw/usb.h"
61 #include "qemu/config-file.h"
62 #include "qemu/error-report.h"
63 #include "trace.h"
64 #include "hw/nmi.h"
65 
66 #include "hw/compat.h"
67 #include "qemu/cutils.h"
68 
69 #include <libfdt.h>
70 
71 /* SLOF memory layout:
72  *
73  * SLOF raw image loaded at 0, copies its romfs right below the flat
74  * device-tree, then position SLOF itself 31M below that
75  *
76  * So we set FW_OVERHEAD to 40MB which should account for all of that
77  * and more
78  *
79  * We load our kernel at 4M, leaving space for SLOF initial image
80  */
81 #define FDT_MAX_SIZE            0x100000
82 #define RTAS_MAX_SIZE           0x10000
83 #define RTAS_MAX_ADDR           0x80000000 /* RTAS must stay below that */
84 #define FW_MAX_SIZE             0x400000
85 #define FW_FILE_NAME            "slof.bin"
86 #define FW_OVERHEAD             0x2800000
87 #define KERNEL_LOAD_ADDR        FW_MAX_SIZE
88 
89 #define MIN_RMA_SLOF            128UL
90 
91 #define TIMEBASE_FREQ           512000000ULL
92 
93 #define PHANDLE_XICP            0x00001111
94 
95 #define HTAB_SIZE(spapr)        (1ULL << ((spapr)->htab_shift))
96 
97 static XICSState *try_create_xics(const char *type, int nr_servers,
98                                   int nr_irqs, Error **errp)
99 {
100     Error *err = NULL;
101     DeviceState *dev;
102 
103     dev = qdev_create(NULL, type);
104     qdev_prop_set_uint32(dev, "nr_servers", nr_servers);
105     qdev_prop_set_uint32(dev, "nr_irqs", nr_irqs);
106     object_property_set_bool(OBJECT(dev), true, "realized", &err);
107     if (err) {
108         error_propagate(errp, err);
109         object_unparent(OBJECT(dev));
110         return NULL;
111     }
112     return XICS_COMMON(dev);
113 }
114 
115 static XICSState *xics_system_init(MachineState *machine,
116                                    int nr_servers, int nr_irqs, Error **errp)
117 {
118     XICSState *icp = NULL;
119 
120     if (kvm_enabled()) {
121         Error *err = NULL;
122 
123         if (machine_kernel_irqchip_allowed(machine)) {
124             icp = try_create_xics(TYPE_KVM_XICS, nr_servers, nr_irqs, &err);
125         }
126         if (machine_kernel_irqchip_required(machine) && !icp) {
127             error_reportf_err(err,
128                               "kernel_irqchip requested but unavailable: ");
129         } else {
130             error_free(err);
131         }
132     }
133 
134     if (!icp) {
135         icp = try_create_xics(TYPE_XICS, nr_servers, nr_irqs, errp);
136     }
137 
138     return icp;
139 }
140 
141 static int spapr_fixup_cpu_smt_dt(void *fdt, int offset, PowerPCCPU *cpu,
142                                   int smt_threads)
143 {
144     int i, ret = 0;
145     uint32_t servers_prop[smt_threads];
146     uint32_t gservers_prop[smt_threads * 2];
147     int index = ppc_get_vcpu_dt_id(cpu);
148 
149     if (cpu->cpu_version) {
150         ret = fdt_setprop_cell(fdt, offset, "cpu-version", cpu->cpu_version);
151         if (ret < 0) {
152             return ret;
153         }
154     }
155 
156     /* Build interrupt servers and gservers properties */
157     for (i = 0; i < smt_threads; i++) {
158         servers_prop[i] = cpu_to_be32(index + i);
159         /* Hack, direct the group queues back to cpu 0 */
160         gservers_prop[i*2] = cpu_to_be32(index + i);
161         gservers_prop[i*2 + 1] = 0;
162     }
163     ret = fdt_setprop(fdt, offset, "ibm,ppc-interrupt-server#s",
164                       servers_prop, sizeof(servers_prop));
165     if (ret < 0) {
166         return ret;
167     }
168     ret = fdt_setprop(fdt, offset, "ibm,ppc-interrupt-gserver#s",
169                       gservers_prop, sizeof(gservers_prop));
170 
171     return ret;
172 }
173 
174 static int spapr_fixup_cpu_numa_dt(void *fdt, int offset, CPUState *cs)
175 {
176     int ret = 0;
177     PowerPCCPU *cpu = POWERPC_CPU(cs);
178     int index = ppc_get_vcpu_dt_id(cpu);
179     uint32_t associativity[] = {cpu_to_be32(0x5),
180                                 cpu_to_be32(0x0),
181                                 cpu_to_be32(0x0),
182                                 cpu_to_be32(0x0),
183                                 cpu_to_be32(cs->numa_node),
184                                 cpu_to_be32(index)};
185 
186     /* Advertise NUMA via ibm,associativity */
187     if (nb_numa_nodes > 1) {
188         ret = fdt_setprop(fdt, offset, "ibm,associativity", associativity,
189                           sizeof(associativity));
190     }
191 
192     return ret;
193 }
194 
195 static int spapr_fixup_cpu_dt(void *fdt, sPAPRMachineState *spapr)
196 {
197     int ret = 0, offset, cpus_offset;
198     CPUState *cs;
199     char cpu_model[32];
200     int smt = kvmppc_smt_threads();
201     uint32_t pft_size_prop[] = {0, cpu_to_be32(spapr->htab_shift)};
202 
203     CPU_FOREACH(cs) {
204         PowerPCCPU *cpu = POWERPC_CPU(cs);
205         DeviceClass *dc = DEVICE_GET_CLASS(cs);
206         int index = ppc_get_vcpu_dt_id(cpu);
207 
208         if ((index % smt) != 0) {
209             continue;
210         }
211 
212         snprintf(cpu_model, 32, "%s@%x", dc->fw_name, index);
213 
214         cpus_offset = fdt_path_offset(fdt, "/cpus");
215         if (cpus_offset < 0) {
216             cpus_offset = fdt_add_subnode(fdt, fdt_path_offset(fdt, "/"),
217                                           "cpus");
218             if (cpus_offset < 0) {
219                 return cpus_offset;
220             }
221         }
222         offset = fdt_subnode_offset(fdt, cpus_offset, cpu_model);
223         if (offset < 0) {
224             offset = fdt_add_subnode(fdt, cpus_offset, cpu_model);
225             if (offset < 0) {
226                 return offset;
227             }
228         }
229 
230         ret = fdt_setprop(fdt, offset, "ibm,pft-size",
231                           pft_size_prop, sizeof(pft_size_prop));
232         if (ret < 0) {
233             return ret;
234         }
235 
236         ret = spapr_fixup_cpu_numa_dt(fdt, offset, cs);
237         if (ret < 0) {
238             return ret;
239         }
240 
241         ret = spapr_fixup_cpu_smt_dt(fdt, offset, cpu,
242                                      ppc_get_compat_smt_threads(cpu));
243         if (ret < 0) {
244             return ret;
245         }
246     }
247     return ret;
248 }
249 
250 
251 static size_t create_page_sizes_prop(CPUPPCState *env, uint32_t *prop,
252                                      size_t maxsize)
253 {
254     size_t maxcells = maxsize / sizeof(uint32_t);
255     int i, j, count;
256     uint32_t *p = prop;
257 
258     for (i = 0; i < PPC_PAGE_SIZES_MAX_SZ; i++) {
259         struct ppc_one_seg_page_size *sps = &env->sps.sps[i];
260 
261         if (!sps->page_shift) {
262             break;
263         }
264         for (count = 0; count < PPC_PAGE_SIZES_MAX_SZ; count++) {
265             if (sps->enc[count].page_shift == 0) {
266                 break;
267             }
268         }
269         if ((p - prop) >= (maxcells - 3 - count * 2)) {
270             break;
271         }
272         *(p++) = cpu_to_be32(sps->page_shift);
273         *(p++) = cpu_to_be32(sps->slb_enc);
274         *(p++) = cpu_to_be32(count);
275         for (j = 0; j < count; j++) {
276             *(p++) = cpu_to_be32(sps->enc[j].page_shift);
277             *(p++) = cpu_to_be32(sps->enc[j].pte_enc);
278         }
279     }
280 
281     return (p - prop) * sizeof(uint32_t);
282 }
283 
284 static hwaddr spapr_node0_size(void)
285 {
286     MachineState *machine = MACHINE(qdev_get_machine());
287 
288     if (nb_numa_nodes) {
289         int i;
290         for (i = 0; i < nb_numa_nodes; ++i) {
291             if (numa_info[i].node_mem) {
292                 return MIN(pow2floor(numa_info[i].node_mem),
293                            machine->ram_size);
294             }
295         }
296     }
297     return machine->ram_size;
298 }
299 
300 #define _FDT(exp) \
301     do { \
302         int ret = (exp);                                           \
303         if (ret < 0) {                                             \
304             fprintf(stderr, "qemu: error creating device tree: %s: %s\n", \
305                     #exp, fdt_strerror(ret));                      \
306             exit(1);                                               \
307         }                                                          \
308     } while (0)
309 
310 static void add_str(GString *s, const gchar *s1)
311 {
312     g_string_append_len(s, s1, strlen(s1) + 1);
313 }
314 
315 static void *spapr_create_fdt_skel(hwaddr initrd_base,
316                                    hwaddr initrd_size,
317                                    hwaddr kernel_size,
318                                    bool little_endian,
319                                    const char *kernel_cmdline,
320                                    uint32_t epow_irq)
321 {
322     void *fdt;
323     uint32_t start_prop = cpu_to_be32(initrd_base);
324     uint32_t end_prop = cpu_to_be32(initrd_base + initrd_size);
325     GString *hypertas = g_string_sized_new(256);
326     GString *qemu_hypertas = g_string_sized_new(256);
327     uint32_t refpoints[] = {cpu_to_be32(0x4), cpu_to_be32(0x4)};
328     uint32_t interrupt_server_ranges_prop[] = {0, cpu_to_be32(max_cpus)};
329     unsigned char vec5[] = {0x0, 0x0, 0x0, 0x0, 0x0, 0x80};
330     char *buf;
331 
332     add_str(hypertas, "hcall-pft");
333     add_str(hypertas, "hcall-term");
334     add_str(hypertas, "hcall-dabr");
335     add_str(hypertas, "hcall-interrupt");
336     add_str(hypertas, "hcall-tce");
337     add_str(hypertas, "hcall-vio");
338     add_str(hypertas, "hcall-splpar");
339     add_str(hypertas, "hcall-bulk");
340     add_str(hypertas, "hcall-set-mode");
341     add_str(qemu_hypertas, "hcall-memop1");
342 
343     fdt = g_malloc0(FDT_MAX_SIZE);
344     _FDT((fdt_create(fdt, FDT_MAX_SIZE)));
345 
346     if (kernel_size) {
347         _FDT((fdt_add_reservemap_entry(fdt, KERNEL_LOAD_ADDR, kernel_size)));
348     }
349     if (initrd_size) {
350         _FDT((fdt_add_reservemap_entry(fdt, initrd_base, initrd_size)));
351     }
352     _FDT((fdt_finish_reservemap(fdt)));
353 
354     /* Root node */
355     _FDT((fdt_begin_node(fdt, "")));
356     _FDT((fdt_property_string(fdt, "device_type", "chrp")));
357     _FDT((fdt_property_string(fdt, "model", "IBM pSeries (emulated by qemu)")));
358     _FDT((fdt_property_string(fdt, "compatible", "qemu,pseries")));
359 
360     /*
361      * Add info to guest to indentify which host is it being run on
362      * and what is the uuid of the guest
363      */
364     if (kvmppc_get_host_model(&buf)) {
365         _FDT((fdt_property_string(fdt, "host-model", buf)));
366         g_free(buf);
367     }
368     if (kvmppc_get_host_serial(&buf)) {
369         _FDT((fdt_property_string(fdt, "host-serial", buf)));
370         g_free(buf);
371     }
372 
373     buf = g_strdup_printf(UUID_FMT, qemu_uuid[0], qemu_uuid[1],
374                           qemu_uuid[2], qemu_uuid[3], qemu_uuid[4],
375                           qemu_uuid[5], qemu_uuid[6], qemu_uuid[7],
376                           qemu_uuid[8], qemu_uuid[9], qemu_uuid[10],
377                           qemu_uuid[11], qemu_uuid[12], qemu_uuid[13],
378                           qemu_uuid[14], qemu_uuid[15]);
379 
380     _FDT((fdt_property_string(fdt, "vm,uuid", buf)));
381     if (qemu_uuid_set) {
382         _FDT((fdt_property_string(fdt, "system-id", buf)));
383     }
384     g_free(buf);
385 
386     if (qemu_get_vm_name()) {
387         _FDT((fdt_property_string(fdt, "ibm,partition-name",
388                                   qemu_get_vm_name())));
389     }
390 
391     _FDT((fdt_property_cell(fdt, "#address-cells", 0x2)));
392     _FDT((fdt_property_cell(fdt, "#size-cells", 0x2)));
393 
394     /* /chosen */
395     _FDT((fdt_begin_node(fdt, "chosen")));
396 
397     /* Set Form1_affinity */
398     _FDT((fdt_property(fdt, "ibm,architecture-vec-5", vec5, sizeof(vec5))));
399 
400     _FDT((fdt_property_string(fdt, "bootargs", kernel_cmdline)));
401     _FDT((fdt_property(fdt, "linux,initrd-start",
402                        &start_prop, sizeof(start_prop))));
403     _FDT((fdt_property(fdt, "linux,initrd-end",
404                        &end_prop, sizeof(end_prop))));
405     if (kernel_size) {
406         uint64_t kprop[2] = { cpu_to_be64(KERNEL_LOAD_ADDR),
407                               cpu_to_be64(kernel_size) };
408 
409         _FDT((fdt_property(fdt, "qemu,boot-kernel", &kprop, sizeof(kprop))));
410         if (little_endian) {
411             _FDT((fdt_property(fdt, "qemu,boot-kernel-le", NULL, 0)));
412         }
413     }
414     if (boot_menu) {
415         _FDT((fdt_property_cell(fdt, "qemu,boot-menu", boot_menu)));
416     }
417     _FDT((fdt_property_cell(fdt, "qemu,graphic-width", graphic_width)));
418     _FDT((fdt_property_cell(fdt, "qemu,graphic-height", graphic_height)));
419     _FDT((fdt_property_cell(fdt, "qemu,graphic-depth", graphic_depth)));
420 
421     _FDT((fdt_end_node(fdt)));
422 
423     /* RTAS */
424     _FDT((fdt_begin_node(fdt, "rtas")));
425 
426     if (!kvm_enabled() || kvmppc_spapr_use_multitce()) {
427         add_str(hypertas, "hcall-multi-tce");
428     }
429     _FDT((fdt_property(fdt, "ibm,hypertas-functions", hypertas->str,
430                        hypertas->len)));
431     g_string_free(hypertas, TRUE);
432     _FDT((fdt_property(fdt, "qemu,hypertas-functions", qemu_hypertas->str,
433                        qemu_hypertas->len)));
434     g_string_free(qemu_hypertas, TRUE);
435 
436     _FDT((fdt_property(fdt, "ibm,associativity-reference-points",
437         refpoints, sizeof(refpoints))));
438 
439     _FDT((fdt_property_cell(fdt, "rtas-error-log-max", RTAS_ERROR_LOG_MAX)));
440     _FDT((fdt_property_cell(fdt, "rtas-event-scan-rate",
441                             RTAS_EVENT_SCAN_RATE)));
442 
443     if (msi_nonbroken) {
444         _FDT((fdt_property(fdt, "ibm,change-msix-capable", NULL, 0)));
445     }
446 
447     /*
448      * According to PAPR, rtas ibm,os-term does not guarantee a return
449      * back to the guest cpu.
450      *
451      * While an additional ibm,extended-os-term property indicates that
452      * rtas call return will always occur. Set this property.
453      */
454     _FDT((fdt_property(fdt, "ibm,extended-os-term", NULL, 0)));
455 
456     _FDT((fdt_end_node(fdt)));
457 
458     /* interrupt controller */
459     _FDT((fdt_begin_node(fdt, "interrupt-controller")));
460 
461     _FDT((fdt_property_string(fdt, "device_type",
462                               "PowerPC-External-Interrupt-Presentation")));
463     _FDT((fdt_property_string(fdt, "compatible", "IBM,ppc-xicp")));
464     _FDT((fdt_property(fdt, "interrupt-controller", NULL, 0)));
465     _FDT((fdt_property(fdt, "ibm,interrupt-server-ranges",
466                        interrupt_server_ranges_prop,
467                        sizeof(interrupt_server_ranges_prop))));
468     _FDT((fdt_property_cell(fdt, "#interrupt-cells", 2)));
469     _FDT((fdt_property_cell(fdt, "linux,phandle", PHANDLE_XICP)));
470     _FDT((fdt_property_cell(fdt, "phandle", PHANDLE_XICP)));
471 
472     _FDT((fdt_end_node(fdt)));
473 
474     /* vdevice */
475     _FDT((fdt_begin_node(fdt, "vdevice")));
476 
477     _FDT((fdt_property_string(fdt, "device_type", "vdevice")));
478     _FDT((fdt_property_string(fdt, "compatible", "IBM,vdevice")));
479     _FDT((fdt_property_cell(fdt, "#address-cells", 0x1)));
480     _FDT((fdt_property_cell(fdt, "#size-cells", 0x0)));
481     _FDT((fdt_property_cell(fdt, "#interrupt-cells", 0x2)));
482     _FDT((fdt_property(fdt, "interrupt-controller", NULL, 0)));
483 
484     _FDT((fdt_end_node(fdt)));
485 
486     /* event-sources */
487     spapr_events_fdt_skel(fdt, epow_irq);
488 
489     /* /hypervisor node */
490     if (kvm_enabled()) {
491         uint8_t hypercall[16];
492 
493         /* indicate KVM hypercall interface */
494         _FDT((fdt_begin_node(fdt, "hypervisor")));
495         _FDT((fdt_property_string(fdt, "compatible", "linux,kvm")));
496         if (kvmppc_has_cap_fixup_hcalls()) {
497             /*
498              * Older KVM versions with older guest kernels were broken with the
499              * magic page, don't allow the guest to map it.
500              */
501             if (!kvmppc_get_hypercall(first_cpu->env_ptr, hypercall,
502                                       sizeof(hypercall))) {
503                 _FDT((fdt_property(fdt, "hcall-instructions", hypercall,
504                                    sizeof(hypercall))));
505             }
506         }
507         _FDT((fdt_end_node(fdt)));
508     }
509 
510     _FDT((fdt_end_node(fdt))); /* close root node */
511     _FDT((fdt_finish(fdt)));
512 
513     return fdt;
514 }
515 
516 static int spapr_populate_memory_node(void *fdt, int nodeid, hwaddr start,
517                                        hwaddr size)
518 {
519     uint32_t associativity[] = {
520         cpu_to_be32(0x4), /* length */
521         cpu_to_be32(0x0), cpu_to_be32(0x0),
522         cpu_to_be32(0x0), cpu_to_be32(nodeid)
523     };
524     char mem_name[32];
525     uint64_t mem_reg_property[2];
526     int off;
527 
528     mem_reg_property[0] = cpu_to_be64(start);
529     mem_reg_property[1] = cpu_to_be64(size);
530 
531     sprintf(mem_name, "memory@" TARGET_FMT_lx, start);
532     off = fdt_add_subnode(fdt, 0, mem_name);
533     _FDT(off);
534     _FDT((fdt_setprop_string(fdt, off, "device_type", "memory")));
535     _FDT((fdt_setprop(fdt, off, "reg", mem_reg_property,
536                       sizeof(mem_reg_property))));
537     _FDT((fdt_setprop(fdt, off, "ibm,associativity", associativity,
538                       sizeof(associativity))));
539     return off;
540 }
541 
542 static int spapr_populate_memory(sPAPRMachineState *spapr, void *fdt)
543 {
544     MachineState *machine = MACHINE(spapr);
545     hwaddr mem_start, node_size;
546     int i, nb_nodes = nb_numa_nodes;
547     NodeInfo *nodes = numa_info;
548     NodeInfo ramnode;
549 
550     /* No NUMA nodes, assume there is just one node with whole RAM */
551     if (!nb_numa_nodes) {
552         nb_nodes = 1;
553         ramnode.node_mem = machine->ram_size;
554         nodes = &ramnode;
555     }
556 
557     for (i = 0, mem_start = 0; i < nb_nodes; ++i) {
558         if (!nodes[i].node_mem) {
559             continue;
560         }
561         if (mem_start >= machine->ram_size) {
562             node_size = 0;
563         } else {
564             node_size = nodes[i].node_mem;
565             if (node_size > machine->ram_size - mem_start) {
566                 node_size = machine->ram_size - mem_start;
567             }
568         }
569         if (!mem_start) {
570             /* ppc_spapr_init() checks for rma_size <= node0_size already */
571             spapr_populate_memory_node(fdt, i, 0, spapr->rma_size);
572             mem_start += spapr->rma_size;
573             node_size -= spapr->rma_size;
574         }
575         for ( ; node_size; ) {
576             hwaddr sizetmp = pow2floor(node_size);
577 
578             /* mem_start != 0 here */
579             if (ctzl(mem_start) < ctzl(sizetmp)) {
580                 sizetmp = 1ULL << ctzl(mem_start);
581             }
582 
583             spapr_populate_memory_node(fdt, i, mem_start, sizetmp);
584             node_size -= sizetmp;
585             mem_start += sizetmp;
586         }
587     }
588 
589     return 0;
590 }
591 
592 static void spapr_populate_cpu_dt(CPUState *cs, void *fdt, int offset,
593                                   sPAPRMachineState *spapr)
594 {
595     PowerPCCPU *cpu = POWERPC_CPU(cs);
596     CPUPPCState *env = &cpu->env;
597     PowerPCCPUClass *pcc = POWERPC_CPU_GET_CLASS(cs);
598     int index = ppc_get_vcpu_dt_id(cpu);
599     uint32_t segs[] = {cpu_to_be32(28), cpu_to_be32(40),
600                        0xffffffff, 0xffffffff};
601     uint32_t tbfreq = kvm_enabled() ? kvmppc_get_tbfreq() : TIMEBASE_FREQ;
602     uint32_t cpufreq = kvm_enabled() ? kvmppc_get_clockfreq() : 1000000000;
603     uint32_t page_sizes_prop[64];
604     size_t page_sizes_prop_size;
605     uint32_t vcpus_per_socket = smp_threads * smp_cores;
606     uint32_t pft_size_prop[] = {0, cpu_to_be32(spapr->htab_shift)};
607 
608     /* Note: we keep CI large pages off for now because a 64K capable guest
609      * provisioned with large pages might otherwise try to map a qemu
610      * framebuffer (or other kind of memory mapped PCI BAR) using 64K pages
611      * even if that qemu runs on a 4k host.
612      *
613      * We can later add this bit back when we are confident this is not
614      * an issue (!HV KVM or 64K host)
615      */
616     uint8_t pa_features_206[] = { 6, 0,
617         0xf6, 0x1f, 0xc7, 0x00, 0x80, 0xc0 };
618     uint8_t pa_features_207[] = { 24, 0,
619         0xf6, 0x1f, 0xc7, 0xc0, 0x80, 0xf0,
620         0x80, 0x00, 0x00, 0x00, 0x00, 0x00,
621         0x00, 0x00, 0x00, 0x00, 0x80, 0x00,
622         0x80, 0x00, 0x80, 0x00, 0x80, 0x00 };
623     uint8_t *pa_features;
624     size_t pa_size;
625 
626     _FDT((fdt_setprop_cell(fdt, offset, "reg", index)));
627     _FDT((fdt_setprop_string(fdt, offset, "device_type", "cpu")));
628 
629     _FDT((fdt_setprop_cell(fdt, offset, "cpu-version", env->spr[SPR_PVR])));
630     _FDT((fdt_setprop_cell(fdt, offset, "d-cache-block-size",
631                            env->dcache_line_size)));
632     _FDT((fdt_setprop_cell(fdt, offset, "d-cache-line-size",
633                            env->dcache_line_size)));
634     _FDT((fdt_setprop_cell(fdt, offset, "i-cache-block-size",
635                            env->icache_line_size)));
636     _FDT((fdt_setprop_cell(fdt, offset, "i-cache-line-size",
637                            env->icache_line_size)));
638 
639     if (pcc->l1_dcache_size) {
640         _FDT((fdt_setprop_cell(fdt, offset, "d-cache-size",
641                                pcc->l1_dcache_size)));
642     } else {
643         fprintf(stderr, "Warning: Unknown L1 dcache size for cpu\n");
644     }
645     if (pcc->l1_icache_size) {
646         _FDT((fdt_setprop_cell(fdt, offset, "i-cache-size",
647                                pcc->l1_icache_size)));
648     } else {
649         fprintf(stderr, "Warning: Unknown L1 icache size for cpu\n");
650     }
651 
652     _FDT((fdt_setprop_cell(fdt, offset, "timebase-frequency", tbfreq)));
653     _FDT((fdt_setprop_cell(fdt, offset, "clock-frequency", cpufreq)));
654     _FDT((fdt_setprop_cell(fdt, offset, "slb-size", env->slb_nr)));
655     _FDT((fdt_setprop_cell(fdt, offset, "ibm,slb-size", env->slb_nr)));
656     _FDT((fdt_setprop_string(fdt, offset, "status", "okay")));
657     _FDT((fdt_setprop(fdt, offset, "64-bit", NULL, 0)));
658 
659     if (env->spr_cb[SPR_PURR].oea_read) {
660         _FDT((fdt_setprop(fdt, offset, "ibm,purr", NULL, 0)));
661     }
662 
663     if (env->mmu_model & POWERPC_MMU_1TSEG) {
664         _FDT((fdt_setprop(fdt, offset, "ibm,processor-segment-sizes",
665                           segs, sizeof(segs))));
666     }
667 
668     /* Advertise VMX/VSX (vector extensions) if available
669      *   0 / no property == no vector extensions
670      *   1               == VMX / Altivec available
671      *   2               == VSX available */
672     if (env->insns_flags & PPC_ALTIVEC) {
673         uint32_t vmx = (env->insns_flags2 & PPC2_VSX) ? 2 : 1;
674 
675         _FDT((fdt_setprop_cell(fdt, offset, "ibm,vmx", vmx)));
676     }
677 
678     /* Advertise DFP (Decimal Floating Point) if available
679      *   0 / no property == no DFP
680      *   1               == DFP available */
681     if (env->insns_flags2 & PPC2_DFP) {
682         _FDT((fdt_setprop_cell(fdt, offset, "ibm,dfp", 1)));
683     }
684 
685     page_sizes_prop_size = create_page_sizes_prop(env, page_sizes_prop,
686                                                   sizeof(page_sizes_prop));
687     if (page_sizes_prop_size) {
688         _FDT((fdt_setprop(fdt, offset, "ibm,segment-page-sizes",
689                           page_sizes_prop, page_sizes_prop_size)));
690     }
691 
692     /* Do the ibm,pa-features property, adjust it for ci-large-pages */
693     if (env->mmu_model == POWERPC_MMU_2_06) {
694         pa_features = pa_features_206;
695         pa_size = sizeof(pa_features_206);
696     } else /* env->mmu_model == POWERPC_MMU_2_07 */ {
697         pa_features = pa_features_207;
698         pa_size = sizeof(pa_features_207);
699     }
700     if (env->ci_large_pages) {
701         pa_features[3] |= 0x20;
702     }
703     _FDT((fdt_setprop(fdt, offset, "ibm,pa-features", pa_features, pa_size)));
704 
705     _FDT((fdt_setprop_cell(fdt, offset, "ibm,chip-id",
706                            cs->cpu_index / vcpus_per_socket)));
707 
708     _FDT((fdt_setprop(fdt, offset, "ibm,pft-size",
709                       pft_size_prop, sizeof(pft_size_prop))));
710 
711     _FDT(spapr_fixup_cpu_numa_dt(fdt, offset, cs));
712 
713     _FDT(spapr_fixup_cpu_smt_dt(fdt, offset, cpu,
714                                 ppc_get_compat_smt_threads(cpu)));
715 }
716 
717 static void spapr_populate_cpus_dt_node(void *fdt, sPAPRMachineState *spapr)
718 {
719     CPUState *cs;
720     int cpus_offset;
721     char *nodename;
722     int smt = kvmppc_smt_threads();
723 
724     cpus_offset = fdt_add_subnode(fdt, 0, "cpus");
725     _FDT(cpus_offset);
726     _FDT((fdt_setprop_cell(fdt, cpus_offset, "#address-cells", 0x1)));
727     _FDT((fdt_setprop_cell(fdt, cpus_offset, "#size-cells", 0x0)));
728 
729     /*
730      * We walk the CPUs in reverse order to ensure that CPU DT nodes
731      * created by fdt_add_subnode() end up in the right order in FDT
732      * for the guest kernel the enumerate the CPUs correctly.
733      */
734     CPU_FOREACH_REVERSE(cs) {
735         PowerPCCPU *cpu = POWERPC_CPU(cs);
736         int index = ppc_get_vcpu_dt_id(cpu);
737         DeviceClass *dc = DEVICE_GET_CLASS(cs);
738         int offset;
739 
740         if ((index % smt) != 0) {
741             continue;
742         }
743 
744         nodename = g_strdup_printf("%s@%x", dc->fw_name, index);
745         offset = fdt_add_subnode(fdt, cpus_offset, nodename);
746         g_free(nodename);
747         _FDT(offset);
748         spapr_populate_cpu_dt(cs, fdt, offset, spapr);
749     }
750 
751 }
752 
753 /*
754  * Adds ibm,dynamic-reconfiguration-memory node.
755  * Refer to docs/specs/ppc-spapr-hotplug.txt for the documentation
756  * of this device tree node.
757  */
758 static int spapr_populate_drconf_memory(sPAPRMachineState *spapr, void *fdt)
759 {
760     MachineState *machine = MACHINE(spapr);
761     int ret, i, offset;
762     uint64_t lmb_size = SPAPR_MEMORY_BLOCK_SIZE;
763     uint32_t prop_lmb_size[] = {0, cpu_to_be32(lmb_size)};
764     uint32_t nr_lmbs = (machine->maxram_size - machine->ram_size)/lmb_size;
765     uint32_t *int_buf, *cur_index, buf_len;
766     int nr_nodes = nb_numa_nodes ? nb_numa_nodes : 1;
767 
768     /*
769      * Don't create the node if there are no DR LMBs.
770      */
771     if (!nr_lmbs) {
772         return 0;
773     }
774 
775     /*
776      * Allocate enough buffer size to fit in ibm,dynamic-memory
777      * or ibm,associativity-lookup-arrays
778      */
779     buf_len = MAX(nr_lmbs * SPAPR_DR_LMB_LIST_ENTRY_SIZE + 1, nr_nodes * 4 + 2)
780               * sizeof(uint32_t);
781     cur_index = int_buf = g_malloc0(buf_len);
782 
783     offset = fdt_add_subnode(fdt, 0, "ibm,dynamic-reconfiguration-memory");
784 
785     ret = fdt_setprop(fdt, offset, "ibm,lmb-size", prop_lmb_size,
786                     sizeof(prop_lmb_size));
787     if (ret < 0) {
788         goto out;
789     }
790 
791     ret = fdt_setprop_cell(fdt, offset, "ibm,memory-flags-mask", 0xff);
792     if (ret < 0) {
793         goto out;
794     }
795 
796     ret = fdt_setprop_cell(fdt, offset, "ibm,memory-preservation-time", 0x0);
797     if (ret < 0) {
798         goto out;
799     }
800 
801     /* ibm,dynamic-memory */
802     int_buf[0] = cpu_to_be32(nr_lmbs);
803     cur_index++;
804     for (i = 0; i < nr_lmbs; i++) {
805         sPAPRDRConnector *drc;
806         sPAPRDRConnectorClass *drck;
807         uint64_t addr = i * lmb_size + spapr->hotplug_memory.base;;
808         uint32_t *dynamic_memory = cur_index;
809 
810         drc = spapr_dr_connector_by_id(SPAPR_DR_CONNECTOR_TYPE_LMB,
811                                        addr/lmb_size);
812         g_assert(drc);
813         drck = SPAPR_DR_CONNECTOR_GET_CLASS(drc);
814 
815         dynamic_memory[0] = cpu_to_be32(addr >> 32);
816         dynamic_memory[1] = cpu_to_be32(addr & 0xffffffff);
817         dynamic_memory[2] = cpu_to_be32(drck->get_index(drc));
818         dynamic_memory[3] = cpu_to_be32(0); /* reserved */
819         dynamic_memory[4] = cpu_to_be32(numa_get_node(addr, NULL));
820         if (addr < machine->ram_size ||
821                     memory_region_present(get_system_memory(), addr)) {
822             dynamic_memory[5] = cpu_to_be32(SPAPR_LMB_FLAGS_ASSIGNED);
823         } else {
824             dynamic_memory[5] = cpu_to_be32(0);
825         }
826 
827         cur_index += SPAPR_DR_LMB_LIST_ENTRY_SIZE;
828     }
829     ret = fdt_setprop(fdt, offset, "ibm,dynamic-memory", int_buf, buf_len);
830     if (ret < 0) {
831         goto out;
832     }
833 
834     /* ibm,associativity-lookup-arrays */
835     cur_index = int_buf;
836     int_buf[0] = cpu_to_be32(nr_nodes);
837     int_buf[1] = cpu_to_be32(4); /* Number of entries per associativity list */
838     cur_index += 2;
839     for (i = 0; i < nr_nodes; i++) {
840         uint32_t associativity[] = {
841             cpu_to_be32(0x0),
842             cpu_to_be32(0x0),
843             cpu_to_be32(0x0),
844             cpu_to_be32(i)
845         };
846         memcpy(cur_index, associativity, sizeof(associativity));
847         cur_index += 4;
848     }
849     ret = fdt_setprop(fdt, offset, "ibm,associativity-lookup-arrays", int_buf,
850             (cur_index - int_buf) * sizeof(uint32_t));
851 out:
852     g_free(int_buf);
853     return ret;
854 }
855 
856 int spapr_h_cas_compose_response(sPAPRMachineState *spapr,
857                                  target_ulong addr, target_ulong size,
858                                  bool cpu_update, bool memory_update)
859 {
860     void *fdt, *fdt_skel;
861     sPAPRDeviceTreeUpdateHeader hdr = { .version_id = 1 };
862     sPAPRMachineClass *smc = SPAPR_MACHINE_GET_CLASS(qdev_get_machine());
863 
864     size -= sizeof(hdr);
865 
866     /* Create sceleton */
867     fdt_skel = g_malloc0(size);
868     _FDT((fdt_create(fdt_skel, size)));
869     _FDT((fdt_begin_node(fdt_skel, "")));
870     _FDT((fdt_end_node(fdt_skel)));
871     _FDT((fdt_finish(fdt_skel)));
872     fdt = g_malloc0(size);
873     _FDT((fdt_open_into(fdt_skel, fdt, size)));
874     g_free(fdt_skel);
875 
876     /* Fixup cpu nodes */
877     if (cpu_update) {
878         _FDT((spapr_fixup_cpu_dt(fdt, spapr)));
879     }
880 
881     /* Generate ibm,dynamic-reconfiguration-memory node if required */
882     if (memory_update && smc->dr_lmb_enabled) {
883         _FDT((spapr_populate_drconf_memory(spapr, fdt)));
884     }
885 
886     /* Pack resulting tree */
887     _FDT((fdt_pack(fdt)));
888 
889     if (fdt_totalsize(fdt) + sizeof(hdr) > size) {
890         trace_spapr_cas_failed(size);
891         return -1;
892     }
893 
894     cpu_physical_memory_write(addr, &hdr, sizeof(hdr));
895     cpu_physical_memory_write(addr + sizeof(hdr), fdt, fdt_totalsize(fdt));
896     trace_spapr_cas_continue(fdt_totalsize(fdt) + sizeof(hdr));
897     g_free(fdt);
898 
899     return 0;
900 }
901 
902 static void spapr_finalize_fdt(sPAPRMachineState *spapr,
903                                hwaddr fdt_addr,
904                                hwaddr rtas_addr,
905                                hwaddr rtas_size)
906 {
907     MachineState *machine = MACHINE(qdev_get_machine());
908     sPAPRMachineClass *smc = SPAPR_MACHINE_GET_CLASS(machine);
909     const char *boot_device = machine->boot_order;
910     int ret, i;
911     size_t cb = 0;
912     char *bootlist;
913     void *fdt;
914     sPAPRPHBState *phb;
915 
916     fdt = g_malloc(FDT_MAX_SIZE);
917 
918     /* open out the base tree into a temp buffer for the final tweaks */
919     _FDT((fdt_open_into(spapr->fdt_skel, fdt, FDT_MAX_SIZE)));
920 
921     ret = spapr_populate_memory(spapr, fdt);
922     if (ret < 0) {
923         fprintf(stderr, "couldn't setup memory nodes in fdt\n");
924         exit(1);
925     }
926 
927     ret = spapr_populate_vdevice(spapr->vio_bus, fdt);
928     if (ret < 0) {
929         fprintf(stderr, "couldn't setup vio devices in fdt\n");
930         exit(1);
931     }
932 
933     if (object_resolve_path_type("", TYPE_SPAPR_RNG, NULL)) {
934         ret = spapr_rng_populate_dt(fdt);
935         if (ret < 0) {
936             fprintf(stderr, "could not set up rng device in the fdt\n");
937             exit(1);
938         }
939     }
940 
941     QLIST_FOREACH(phb, &spapr->phbs, list) {
942         ret = spapr_populate_pci_dt(phb, PHANDLE_XICP, fdt);
943     }
944 
945     if (ret < 0) {
946         fprintf(stderr, "couldn't setup PCI devices in fdt\n");
947         exit(1);
948     }
949 
950     /* RTAS */
951     ret = spapr_rtas_device_tree_setup(fdt, rtas_addr, rtas_size);
952     if (ret < 0) {
953         fprintf(stderr, "Couldn't set up RTAS device tree properties\n");
954     }
955 
956     /* cpus */
957     spapr_populate_cpus_dt_node(fdt, spapr);
958 
959     bootlist = get_boot_devices_list(&cb, true);
960     if (cb && bootlist) {
961         int offset = fdt_path_offset(fdt, "/chosen");
962         if (offset < 0) {
963             exit(1);
964         }
965         for (i = 0; i < cb; i++) {
966             if (bootlist[i] == '\n') {
967                 bootlist[i] = ' ';
968             }
969 
970         }
971         ret = fdt_setprop_string(fdt, offset, "qemu,boot-list", bootlist);
972     }
973 
974     if (boot_device && strlen(boot_device)) {
975         int offset = fdt_path_offset(fdt, "/chosen");
976 
977         if (offset < 0) {
978             exit(1);
979         }
980         fdt_setprop_string(fdt, offset, "qemu,boot-device", boot_device);
981     }
982 
983     if (!spapr->has_graphics) {
984         spapr_populate_chosen_stdout(fdt, spapr->vio_bus);
985     }
986 
987     if (smc->dr_lmb_enabled) {
988         _FDT(spapr_drc_populate_dt(fdt, 0, NULL, SPAPR_DR_CONNECTOR_TYPE_LMB));
989     }
990 
991     _FDT((fdt_pack(fdt)));
992 
993     if (fdt_totalsize(fdt) > FDT_MAX_SIZE) {
994         error_report("FDT too big ! 0x%x bytes (max is 0x%x)",
995                      fdt_totalsize(fdt), FDT_MAX_SIZE);
996         exit(1);
997     }
998 
999     qemu_fdt_dumpdtb(fdt, fdt_totalsize(fdt));
1000     cpu_physical_memory_write(fdt_addr, fdt, fdt_totalsize(fdt));
1001 
1002     g_free(bootlist);
1003     g_free(fdt);
1004 }
1005 
1006 static uint64_t translate_kernel_address(void *opaque, uint64_t addr)
1007 {
1008     return (addr & 0x0fffffff) + KERNEL_LOAD_ADDR;
1009 }
1010 
1011 static void emulate_spapr_hypercall(PowerPCCPU *cpu)
1012 {
1013     CPUPPCState *env = &cpu->env;
1014 
1015     if (msr_pr) {
1016         hcall_dprintf("Hypercall made with MSR[PR]=1\n");
1017         env->gpr[3] = H_PRIVILEGE;
1018     } else {
1019         env->gpr[3] = spapr_hypercall(cpu, env->gpr[3], &env->gpr[4]);
1020     }
1021 }
1022 
1023 #define HPTE(_table, _i)   (void *)(((uint64_t *)(_table)) + ((_i) * 2))
1024 #define HPTE_VALID(_hpte)  (tswap64(*((uint64_t *)(_hpte))) & HPTE64_V_VALID)
1025 #define HPTE_DIRTY(_hpte)  (tswap64(*((uint64_t *)(_hpte))) & HPTE64_V_HPTE_DIRTY)
1026 #define CLEAN_HPTE(_hpte)  ((*(uint64_t *)(_hpte)) &= tswap64(~HPTE64_V_HPTE_DIRTY))
1027 #define DIRTY_HPTE(_hpte)  ((*(uint64_t *)(_hpte)) |= tswap64(HPTE64_V_HPTE_DIRTY))
1028 
1029 /*
1030  * Get the fd to access the kernel htab, re-opening it if necessary
1031  */
1032 static int get_htab_fd(sPAPRMachineState *spapr)
1033 {
1034     if (spapr->htab_fd >= 0) {
1035         return spapr->htab_fd;
1036     }
1037 
1038     spapr->htab_fd = kvmppc_get_htab_fd(false);
1039     if (spapr->htab_fd < 0) {
1040         error_report("Unable to open fd for reading hash table from KVM: %s",
1041                      strerror(errno));
1042     }
1043 
1044     return spapr->htab_fd;
1045 }
1046 
1047 static void close_htab_fd(sPAPRMachineState *spapr)
1048 {
1049     if (spapr->htab_fd >= 0) {
1050         close(spapr->htab_fd);
1051     }
1052     spapr->htab_fd = -1;
1053 }
1054 
1055 static int spapr_hpt_shift_for_ramsize(uint64_t ramsize)
1056 {
1057     int shift;
1058 
1059     /* We aim for a hash table of size 1/128 the size of RAM (rounded
1060      * up).  The PAPR recommendation is actually 1/64 of RAM size, but
1061      * that's much more than is needed for Linux guests */
1062     shift = ctz64(pow2ceil(ramsize)) - 7;
1063     shift = MAX(shift, 18); /* Minimum architected size */
1064     shift = MIN(shift, 46); /* Maximum architected size */
1065     return shift;
1066 }
1067 
1068 static void spapr_reallocate_hpt(sPAPRMachineState *spapr, int shift,
1069                                  Error **errp)
1070 {
1071     long rc;
1072 
1073     /* Clean up any HPT info from a previous boot */
1074     g_free(spapr->htab);
1075     spapr->htab = NULL;
1076     spapr->htab_shift = 0;
1077     close_htab_fd(spapr);
1078 
1079     rc = kvmppc_reset_htab(shift);
1080     if (rc < 0) {
1081         /* kernel-side HPT needed, but couldn't allocate one */
1082         error_setg_errno(errp, errno,
1083                          "Failed to allocate KVM HPT of order %d (try smaller maxmem?)",
1084                          shift);
1085         /* This is almost certainly fatal, but if the caller really
1086          * wants to carry on with shift == 0, it's welcome to try */
1087     } else if (rc > 0) {
1088         /* kernel-side HPT allocated */
1089         if (rc != shift) {
1090             error_setg(errp,
1091                        "Requested order %d HPT, but kernel allocated order %ld (try smaller maxmem?)",
1092                        shift, rc);
1093         }
1094 
1095         spapr->htab_shift = shift;
1096         spapr->htab = NULL;
1097     } else {
1098         /* kernel-side HPT not needed, allocate in userspace instead */
1099         size_t size = 1ULL << shift;
1100         int i;
1101 
1102         spapr->htab = qemu_memalign(size, size);
1103         if (!spapr->htab) {
1104             error_setg_errno(errp, errno,
1105                              "Could not allocate HPT of order %d", shift);
1106             return;
1107         }
1108 
1109         memset(spapr->htab, 0, size);
1110         spapr->htab_shift = shift;
1111 
1112         for (i = 0; i < size / HASH_PTE_SIZE_64; i++) {
1113             DIRTY_HPTE(HPTE(spapr->htab, i));
1114         }
1115     }
1116 }
1117 
1118 static int find_unknown_sysbus_device(SysBusDevice *sbdev, void *opaque)
1119 {
1120     bool matched = false;
1121 
1122     if (object_dynamic_cast(OBJECT(sbdev), TYPE_SPAPR_PCI_HOST_BRIDGE)) {
1123         matched = true;
1124     }
1125 
1126     if (!matched) {
1127         error_report("Device %s is not supported by this machine yet.",
1128                      qdev_fw_name(DEVICE(sbdev)));
1129         exit(1);
1130     }
1131 
1132     return 0;
1133 }
1134 
1135 static void ppc_spapr_reset(void)
1136 {
1137     MachineState *machine = MACHINE(qdev_get_machine());
1138     sPAPRMachineState *spapr = SPAPR_MACHINE(machine);
1139     PowerPCCPU *first_ppc_cpu;
1140     uint32_t rtas_limit;
1141 
1142     /* Check for unknown sysbus devices */
1143     foreach_dynamic_sysbus_device(find_unknown_sysbus_device, NULL);
1144 
1145     /* Allocate and/or reset the hash page table */
1146     spapr_reallocate_hpt(spapr,
1147                          spapr_hpt_shift_for_ramsize(machine->maxram_size),
1148                          &error_fatal);
1149 
1150     /* Update the RMA size if necessary */
1151     if (spapr->vrma_adjust) {
1152         spapr->rma_size = kvmppc_rma_size(spapr_node0_size(),
1153                                           spapr->htab_shift);
1154     }
1155 
1156     qemu_devices_reset();
1157 
1158     /*
1159      * We place the device tree and RTAS just below either the top of the RMA,
1160      * or just below 2GB, whichever is lowere, so that it can be
1161      * processed with 32-bit real mode code if necessary
1162      */
1163     rtas_limit = MIN(spapr->rma_size, RTAS_MAX_ADDR);
1164     spapr->rtas_addr = rtas_limit - RTAS_MAX_SIZE;
1165     spapr->fdt_addr = spapr->rtas_addr - FDT_MAX_SIZE;
1166 
1167     /* Load the fdt */
1168     spapr_finalize_fdt(spapr, spapr->fdt_addr, spapr->rtas_addr,
1169                        spapr->rtas_size);
1170 
1171     /* Copy RTAS over */
1172     cpu_physical_memory_write(spapr->rtas_addr, spapr->rtas_blob,
1173                               spapr->rtas_size);
1174 
1175     /* Set up the entry state */
1176     first_ppc_cpu = POWERPC_CPU(first_cpu);
1177     first_ppc_cpu->env.gpr[3] = spapr->fdt_addr;
1178     first_ppc_cpu->env.gpr[5] = 0;
1179     first_cpu->halted = 0;
1180     first_ppc_cpu->env.nip = SPAPR_ENTRY_POINT;
1181 
1182 }
1183 
1184 static void spapr_cpu_reset(void *opaque)
1185 {
1186     sPAPRMachineState *spapr = SPAPR_MACHINE(qdev_get_machine());
1187     PowerPCCPU *cpu = opaque;
1188     CPUState *cs = CPU(cpu);
1189     CPUPPCState *env = &cpu->env;
1190 
1191     cpu_reset(cs);
1192 
1193     /* All CPUs start halted.  CPU0 is unhalted from the machine level
1194      * reset code and the rest are explicitly started up by the guest
1195      * using an RTAS call */
1196     cs->halted = 1;
1197 
1198     env->spr[SPR_HIOR] = 0;
1199 
1200     ppc_hash64_set_external_hpt(cpu, spapr->htab, spapr->htab_shift,
1201                                 &error_fatal);
1202 }
1203 
1204 static void spapr_create_nvram(sPAPRMachineState *spapr)
1205 {
1206     DeviceState *dev = qdev_create(&spapr->vio_bus->bus, "spapr-nvram");
1207     DriveInfo *dinfo = drive_get(IF_PFLASH, 0, 0);
1208 
1209     if (dinfo) {
1210         qdev_prop_set_drive(dev, "drive", blk_by_legacy_dinfo(dinfo),
1211                             &error_fatal);
1212     }
1213 
1214     qdev_init_nofail(dev);
1215 
1216     spapr->nvram = (struct sPAPRNVRAM *)dev;
1217 }
1218 
1219 static void spapr_rtc_create(sPAPRMachineState *spapr)
1220 {
1221     DeviceState *dev = qdev_create(NULL, TYPE_SPAPR_RTC);
1222 
1223     qdev_init_nofail(dev);
1224     spapr->rtc = dev;
1225 
1226     object_property_add_alias(qdev_get_machine(), "rtc-time",
1227                               OBJECT(spapr->rtc), "date", NULL);
1228 }
1229 
1230 /* Returns whether we want to use VGA or not */
1231 static bool spapr_vga_init(PCIBus *pci_bus, Error **errp)
1232 {
1233     switch (vga_interface_type) {
1234     case VGA_NONE:
1235         return false;
1236     case VGA_DEVICE:
1237         return true;
1238     case VGA_STD:
1239     case VGA_VIRTIO:
1240         return pci_vga_init(pci_bus) != NULL;
1241     default:
1242         error_setg(errp,
1243                    "Unsupported VGA mode, only -vga std or -vga virtio is supported");
1244         return false;
1245     }
1246 }
1247 
1248 static int spapr_post_load(void *opaque, int version_id)
1249 {
1250     sPAPRMachineState *spapr = (sPAPRMachineState *)opaque;
1251     int err = 0;
1252 
1253     /* In earlier versions, there was no separate qdev for the PAPR
1254      * RTC, so the RTC offset was stored directly in sPAPREnvironment.
1255      * So when migrating from those versions, poke the incoming offset
1256      * value into the RTC device */
1257     if (version_id < 3) {
1258         err = spapr_rtc_import_offset(spapr->rtc, spapr->rtc_offset);
1259     }
1260 
1261     return err;
1262 }
1263 
1264 static bool version_before_3(void *opaque, int version_id)
1265 {
1266     return version_id < 3;
1267 }
1268 
1269 static const VMStateDescription vmstate_spapr = {
1270     .name = "spapr",
1271     .version_id = 3,
1272     .minimum_version_id = 1,
1273     .post_load = spapr_post_load,
1274     .fields = (VMStateField[]) {
1275         /* used to be @next_irq */
1276         VMSTATE_UNUSED_BUFFER(version_before_3, 0, 4),
1277 
1278         /* RTC offset */
1279         VMSTATE_UINT64_TEST(rtc_offset, sPAPRMachineState, version_before_3),
1280 
1281         VMSTATE_PPC_TIMEBASE_V(tb, sPAPRMachineState, 2),
1282         VMSTATE_END_OF_LIST()
1283     },
1284 };
1285 
1286 static int htab_save_setup(QEMUFile *f, void *opaque)
1287 {
1288     sPAPRMachineState *spapr = opaque;
1289 
1290     /* "Iteration" header */
1291     qemu_put_be32(f, spapr->htab_shift);
1292 
1293     if (spapr->htab) {
1294         spapr->htab_save_index = 0;
1295         spapr->htab_first_pass = true;
1296     } else {
1297         assert(kvm_enabled());
1298     }
1299 
1300 
1301     return 0;
1302 }
1303 
1304 static void htab_save_first_pass(QEMUFile *f, sPAPRMachineState *spapr,
1305                                  int64_t max_ns)
1306 {
1307     bool has_timeout = max_ns != -1;
1308     int htabslots = HTAB_SIZE(spapr) / HASH_PTE_SIZE_64;
1309     int index = spapr->htab_save_index;
1310     int64_t starttime = qemu_clock_get_ns(QEMU_CLOCK_REALTIME);
1311 
1312     assert(spapr->htab_first_pass);
1313 
1314     do {
1315         int chunkstart;
1316 
1317         /* Consume invalid HPTEs */
1318         while ((index < htabslots)
1319                && !HPTE_VALID(HPTE(spapr->htab, index))) {
1320             index++;
1321             CLEAN_HPTE(HPTE(spapr->htab, index));
1322         }
1323 
1324         /* Consume valid HPTEs */
1325         chunkstart = index;
1326         while ((index < htabslots) && (index - chunkstart < USHRT_MAX)
1327                && HPTE_VALID(HPTE(spapr->htab, index))) {
1328             index++;
1329             CLEAN_HPTE(HPTE(spapr->htab, index));
1330         }
1331 
1332         if (index > chunkstart) {
1333             int n_valid = index - chunkstart;
1334 
1335             qemu_put_be32(f, chunkstart);
1336             qemu_put_be16(f, n_valid);
1337             qemu_put_be16(f, 0);
1338             qemu_put_buffer(f, HPTE(spapr->htab, chunkstart),
1339                             HASH_PTE_SIZE_64 * n_valid);
1340 
1341             if (has_timeout &&
1342                 (qemu_clock_get_ns(QEMU_CLOCK_REALTIME) - starttime) > max_ns) {
1343                 break;
1344             }
1345         }
1346     } while ((index < htabslots) && !qemu_file_rate_limit(f));
1347 
1348     if (index >= htabslots) {
1349         assert(index == htabslots);
1350         index = 0;
1351         spapr->htab_first_pass = false;
1352     }
1353     spapr->htab_save_index = index;
1354 }
1355 
1356 static int htab_save_later_pass(QEMUFile *f, sPAPRMachineState *spapr,
1357                                 int64_t max_ns)
1358 {
1359     bool final = max_ns < 0;
1360     int htabslots = HTAB_SIZE(spapr) / HASH_PTE_SIZE_64;
1361     int examined = 0, sent = 0;
1362     int index = spapr->htab_save_index;
1363     int64_t starttime = qemu_clock_get_ns(QEMU_CLOCK_REALTIME);
1364 
1365     assert(!spapr->htab_first_pass);
1366 
1367     do {
1368         int chunkstart, invalidstart;
1369 
1370         /* Consume non-dirty HPTEs */
1371         while ((index < htabslots)
1372                && !HPTE_DIRTY(HPTE(spapr->htab, index))) {
1373             index++;
1374             examined++;
1375         }
1376 
1377         chunkstart = index;
1378         /* Consume valid dirty HPTEs */
1379         while ((index < htabslots) && (index - chunkstart < USHRT_MAX)
1380                && HPTE_DIRTY(HPTE(spapr->htab, index))
1381                && HPTE_VALID(HPTE(spapr->htab, index))) {
1382             CLEAN_HPTE(HPTE(spapr->htab, index));
1383             index++;
1384             examined++;
1385         }
1386 
1387         invalidstart = index;
1388         /* Consume invalid dirty HPTEs */
1389         while ((index < htabslots) && (index - invalidstart < USHRT_MAX)
1390                && HPTE_DIRTY(HPTE(spapr->htab, index))
1391                && !HPTE_VALID(HPTE(spapr->htab, index))) {
1392             CLEAN_HPTE(HPTE(spapr->htab, index));
1393             index++;
1394             examined++;
1395         }
1396 
1397         if (index > chunkstart) {
1398             int n_valid = invalidstart - chunkstart;
1399             int n_invalid = index - invalidstart;
1400 
1401             qemu_put_be32(f, chunkstart);
1402             qemu_put_be16(f, n_valid);
1403             qemu_put_be16(f, n_invalid);
1404             qemu_put_buffer(f, HPTE(spapr->htab, chunkstart),
1405                             HASH_PTE_SIZE_64 * n_valid);
1406             sent += index - chunkstart;
1407 
1408             if (!final && (qemu_clock_get_ns(QEMU_CLOCK_REALTIME) - starttime) > max_ns) {
1409                 break;
1410             }
1411         }
1412 
1413         if (examined >= htabslots) {
1414             break;
1415         }
1416 
1417         if (index >= htabslots) {
1418             assert(index == htabslots);
1419             index = 0;
1420         }
1421     } while ((examined < htabslots) && (!qemu_file_rate_limit(f) || final));
1422 
1423     if (index >= htabslots) {
1424         assert(index == htabslots);
1425         index = 0;
1426     }
1427 
1428     spapr->htab_save_index = index;
1429 
1430     return (examined >= htabslots) && (sent == 0) ? 1 : 0;
1431 }
1432 
1433 #define MAX_ITERATION_NS    5000000 /* 5 ms */
1434 #define MAX_KVM_BUF_SIZE    2048
1435 
1436 static int htab_save_iterate(QEMUFile *f, void *opaque)
1437 {
1438     sPAPRMachineState *spapr = opaque;
1439     int fd;
1440     int rc = 0;
1441 
1442     /* Iteration header */
1443     qemu_put_be32(f, 0);
1444 
1445     if (!spapr->htab) {
1446         assert(kvm_enabled());
1447 
1448         fd = get_htab_fd(spapr);
1449         if (fd < 0) {
1450             return fd;
1451         }
1452 
1453         rc = kvmppc_save_htab(f, fd, MAX_KVM_BUF_SIZE, MAX_ITERATION_NS);
1454         if (rc < 0) {
1455             return rc;
1456         }
1457     } else  if (spapr->htab_first_pass) {
1458         htab_save_first_pass(f, spapr, MAX_ITERATION_NS);
1459     } else {
1460         rc = htab_save_later_pass(f, spapr, MAX_ITERATION_NS);
1461     }
1462 
1463     /* End marker */
1464     qemu_put_be32(f, 0);
1465     qemu_put_be16(f, 0);
1466     qemu_put_be16(f, 0);
1467 
1468     return rc;
1469 }
1470 
1471 static int htab_save_complete(QEMUFile *f, void *opaque)
1472 {
1473     sPAPRMachineState *spapr = opaque;
1474     int fd;
1475 
1476     /* Iteration header */
1477     qemu_put_be32(f, 0);
1478 
1479     if (!spapr->htab) {
1480         int rc;
1481 
1482         assert(kvm_enabled());
1483 
1484         fd = get_htab_fd(spapr);
1485         if (fd < 0) {
1486             return fd;
1487         }
1488 
1489         rc = kvmppc_save_htab(f, fd, MAX_KVM_BUF_SIZE, -1);
1490         if (rc < 0) {
1491             return rc;
1492         }
1493         close_htab_fd(spapr);
1494     } else {
1495         if (spapr->htab_first_pass) {
1496             htab_save_first_pass(f, spapr, -1);
1497         }
1498         htab_save_later_pass(f, spapr, -1);
1499     }
1500 
1501     /* End marker */
1502     qemu_put_be32(f, 0);
1503     qemu_put_be16(f, 0);
1504     qemu_put_be16(f, 0);
1505 
1506     return 0;
1507 }
1508 
1509 static int htab_load(QEMUFile *f, void *opaque, int version_id)
1510 {
1511     sPAPRMachineState *spapr = opaque;
1512     uint32_t section_hdr;
1513     int fd = -1;
1514 
1515     if (version_id < 1 || version_id > 1) {
1516         error_report("htab_load() bad version");
1517         return -EINVAL;
1518     }
1519 
1520     section_hdr = qemu_get_be32(f);
1521 
1522     if (section_hdr) {
1523         Error *local_err = NULL;
1524 
1525         /* First section gives the htab size */
1526         spapr_reallocate_hpt(spapr, section_hdr, &local_err);
1527         if (local_err) {
1528             error_report_err(local_err);
1529             return -EINVAL;
1530         }
1531         return 0;
1532     }
1533 
1534     if (!spapr->htab) {
1535         assert(kvm_enabled());
1536 
1537         fd = kvmppc_get_htab_fd(true);
1538         if (fd < 0) {
1539             error_report("Unable to open fd to restore KVM hash table: %s",
1540                          strerror(errno));
1541         }
1542     }
1543 
1544     while (true) {
1545         uint32_t index;
1546         uint16_t n_valid, n_invalid;
1547 
1548         index = qemu_get_be32(f);
1549         n_valid = qemu_get_be16(f);
1550         n_invalid = qemu_get_be16(f);
1551 
1552         if ((index == 0) && (n_valid == 0) && (n_invalid == 0)) {
1553             /* End of Stream */
1554             break;
1555         }
1556 
1557         if ((index + n_valid + n_invalid) >
1558             (HTAB_SIZE(spapr) / HASH_PTE_SIZE_64)) {
1559             /* Bad index in stream */
1560             error_report(
1561                 "htab_load() bad index %d (%hd+%hd entries) in htab stream (htab_shift=%d)",
1562                 index, n_valid, n_invalid, spapr->htab_shift);
1563             return -EINVAL;
1564         }
1565 
1566         if (spapr->htab) {
1567             if (n_valid) {
1568                 qemu_get_buffer(f, HPTE(spapr->htab, index),
1569                                 HASH_PTE_SIZE_64 * n_valid);
1570             }
1571             if (n_invalid) {
1572                 memset(HPTE(spapr->htab, index + n_valid), 0,
1573                        HASH_PTE_SIZE_64 * n_invalid);
1574             }
1575         } else {
1576             int rc;
1577 
1578             assert(fd >= 0);
1579 
1580             rc = kvmppc_load_htab_chunk(f, fd, index, n_valid, n_invalid);
1581             if (rc < 0) {
1582                 return rc;
1583             }
1584         }
1585     }
1586 
1587     if (!spapr->htab) {
1588         assert(fd >= 0);
1589         close(fd);
1590     }
1591 
1592     return 0;
1593 }
1594 
1595 static SaveVMHandlers savevm_htab_handlers = {
1596     .save_live_setup = htab_save_setup,
1597     .save_live_iterate = htab_save_iterate,
1598     .save_live_complete_precopy = htab_save_complete,
1599     .load_state = htab_load,
1600 };
1601 
1602 static void spapr_boot_set(void *opaque, const char *boot_device,
1603                            Error **errp)
1604 {
1605     MachineState *machine = MACHINE(qdev_get_machine());
1606     machine->boot_order = g_strdup(boot_device);
1607 }
1608 
1609 static void spapr_cpu_init(sPAPRMachineState *spapr, PowerPCCPU *cpu,
1610                            Error **errp)
1611 {
1612     CPUPPCState *env = &cpu->env;
1613 
1614     /* Set time-base frequency to 512 MHz */
1615     cpu_ppc_tb_init(env, TIMEBASE_FREQ);
1616 
1617     /* Enable PAPR mode in TCG or KVM */
1618     cpu_ppc_set_papr(cpu);
1619 
1620     if (cpu->max_compat) {
1621         Error *local_err = NULL;
1622 
1623         ppc_set_compat(cpu, cpu->max_compat, &local_err);
1624         if (local_err) {
1625             error_propagate(errp, local_err);
1626             return;
1627         }
1628     }
1629 
1630     xics_cpu_setup(spapr->icp, cpu);
1631 
1632     qemu_register_reset(spapr_cpu_reset, cpu);
1633 }
1634 
1635 /*
1636  * Reset routine for LMB DR devices.
1637  *
1638  * Unlike PCI DR devices, LMB DR devices explicitly register this reset
1639  * routine. Reset for PCI DR devices will be handled by PHB reset routine
1640  * when it walks all its children devices. LMB devices reset occurs
1641  * as part of spapr_ppc_reset().
1642  */
1643 static void spapr_drc_reset(void *opaque)
1644 {
1645     sPAPRDRConnector *drc = opaque;
1646     DeviceState *d = DEVICE(drc);
1647 
1648     if (d) {
1649         device_reset(d);
1650     }
1651 }
1652 
1653 static void spapr_create_lmb_dr_connectors(sPAPRMachineState *spapr)
1654 {
1655     MachineState *machine = MACHINE(spapr);
1656     uint64_t lmb_size = SPAPR_MEMORY_BLOCK_SIZE;
1657     uint32_t nr_lmbs = (machine->maxram_size - machine->ram_size)/lmb_size;
1658     int i;
1659 
1660     for (i = 0; i < nr_lmbs; i++) {
1661         sPAPRDRConnector *drc;
1662         uint64_t addr;
1663 
1664         addr = i * lmb_size + spapr->hotplug_memory.base;
1665         drc = spapr_dr_connector_new(OBJECT(spapr), SPAPR_DR_CONNECTOR_TYPE_LMB,
1666                                      addr/lmb_size);
1667         qemu_register_reset(spapr_drc_reset, drc);
1668     }
1669 }
1670 
1671 /*
1672  * If RAM size, maxmem size and individual node mem sizes aren't aligned
1673  * to SPAPR_MEMORY_BLOCK_SIZE(256MB), then refuse to start the guest
1674  * since we can't support such unaligned sizes with DRCONF_MEMORY.
1675  */
1676 static void spapr_validate_node_memory(MachineState *machine, Error **errp)
1677 {
1678     int i;
1679 
1680     if (machine->ram_size % SPAPR_MEMORY_BLOCK_SIZE) {
1681         error_setg(errp, "Memory size 0x" RAM_ADDR_FMT
1682                    " is not aligned to %llu MiB",
1683                    machine->ram_size,
1684                    SPAPR_MEMORY_BLOCK_SIZE / M_BYTE);
1685         return;
1686     }
1687 
1688     if (machine->maxram_size % SPAPR_MEMORY_BLOCK_SIZE) {
1689         error_setg(errp, "Maximum memory size 0x" RAM_ADDR_FMT
1690                    " is not aligned to %llu MiB",
1691                    machine->ram_size,
1692                    SPAPR_MEMORY_BLOCK_SIZE / M_BYTE);
1693         return;
1694     }
1695 
1696     for (i = 0; i < nb_numa_nodes; i++) {
1697         if (numa_info[i].node_mem % SPAPR_MEMORY_BLOCK_SIZE) {
1698             error_setg(errp,
1699                        "Node %d memory size 0x%" PRIx64
1700                        " is not aligned to %llu MiB",
1701                        i, numa_info[i].node_mem,
1702                        SPAPR_MEMORY_BLOCK_SIZE / M_BYTE);
1703             return;
1704         }
1705     }
1706 }
1707 
1708 /* pSeries LPAR / sPAPR hardware init */
1709 static void ppc_spapr_init(MachineState *machine)
1710 {
1711     sPAPRMachineState *spapr = SPAPR_MACHINE(machine);
1712     sPAPRMachineClass *smc = SPAPR_MACHINE_GET_CLASS(machine);
1713     const char *kernel_filename = machine->kernel_filename;
1714     const char *kernel_cmdline = machine->kernel_cmdline;
1715     const char *initrd_filename = machine->initrd_filename;
1716     PowerPCCPU *cpu;
1717     PCIHostState *phb;
1718     int i;
1719     MemoryRegion *sysmem = get_system_memory();
1720     MemoryRegion *ram = g_new(MemoryRegion, 1);
1721     MemoryRegion *rma_region;
1722     void *rma = NULL;
1723     hwaddr rma_alloc_size;
1724     hwaddr node0_size = spapr_node0_size();
1725     uint32_t initrd_base = 0;
1726     long kernel_size = 0, initrd_size = 0;
1727     long load_limit, fw_size;
1728     bool kernel_le = false;
1729     char *filename;
1730 
1731     msi_nonbroken = true;
1732 
1733     QLIST_INIT(&spapr->phbs);
1734 
1735     cpu_ppc_hypercall = emulate_spapr_hypercall;
1736 
1737     /* Allocate RMA if necessary */
1738     rma_alloc_size = kvmppc_alloc_rma(&rma);
1739 
1740     if (rma_alloc_size == -1) {
1741         error_report("Unable to create RMA");
1742         exit(1);
1743     }
1744 
1745     if (rma_alloc_size && (rma_alloc_size < node0_size)) {
1746         spapr->rma_size = rma_alloc_size;
1747     } else {
1748         spapr->rma_size = node0_size;
1749 
1750         /* With KVM, we don't actually know whether KVM supports an
1751          * unbounded RMA (PR KVM) or is limited by the hash table size
1752          * (HV KVM using VRMA), so we always assume the latter
1753          *
1754          * In that case, we also limit the initial allocations for RTAS
1755          * etc... to 256M since we have no way to know what the VRMA size
1756          * is going to be as it depends on the size of the hash table
1757          * isn't determined yet.
1758          */
1759         if (kvm_enabled()) {
1760             spapr->vrma_adjust = 1;
1761             spapr->rma_size = MIN(spapr->rma_size, 0x10000000);
1762         }
1763     }
1764 
1765     if (spapr->rma_size > node0_size) {
1766         error_report("Numa node 0 has to span the RMA (%#08"HWADDR_PRIx")",
1767                      spapr->rma_size);
1768         exit(1);
1769     }
1770 
1771     /* Setup a load limit for the ramdisk leaving room for SLOF and FDT */
1772     load_limit = MIN(spapr->rma_size, RTAS_MAX_ADDR) - FW_OVERHEAD;
1773 
1774     /* Set up Interrupt Controller before we create the VCPUs */
1775     spapr->icp = xics_system_init(machine,
1776                                   DIV_ROUND_UP(max_cpus * kvmppc_smt_threads(),
1777                                                smp_threads),
1778                                   XICS_IRQS, &error_fatal);
1779 
1780     if (smc->dr_lmb_enabled) {
1781         spapr_validate_node_memory(machine, &error_fatal);
1782     }
1783 
1784     /* init CPUs */
1785     if (machine->cpu_model == NULL) {
1786         machine->cpu_model = kvm_enabled() ? "host" : "POWER7";
1787     }
1788     for (i = 0; i < smp_cpus; i++) {
1789         cpu = cpu_ppc_init(machine->cpu_model);
1790         if (cpu == NULL) {
1791             error_report("Unable to find PowerPC CPU definition");
1792             exit(1);
1793         }
1794         spapr_cpu_init(spapr, cpu, &error_fatal);
1795     }
1796 
1797     if (kvm_enabled()) {
1798         /* Enable H_LOGICAL_CI_* so SLOF can talk to in-kernel devices */
1799         kvmppc_enable_logical_ci_hcalls();
1800         kvmppc_enable_set_mode_hcall();
1801     }
1802 
1803     /* allocate RAM */
1804     memory_region_allocate_system_memory(ram, NULL, "ppc_spapr.ram",
1805                                          machine->ram_size);
1806     memory_region_add_subregion(sysmem, 0, ram);
1807 
1808     if (rma_alloc_size && rma) {
1809         rma_region = g_new(MemoryRegion, 1);
1810         memory_region_init_ram_ptr(rma_region, NULL, "ppc_spapr.rma",
1811                                    rma_alloc_size, rma);
1812         vmstate_register_ram_global(rma_region);
1813         memory_region_add_subregion(sysmem, 0, rma_region);
1814     }
1815 
1816     /* initialize hotplug memory address space */
1817     if (machine->ram_size < machine->maxram_size) {
1818         ram_addr_t hotplug_mem_size = machine->maxram_size - machine->ram_size;
1819 
1820         if (machine->ram_slots > SPAPR_MAX_RAM_SLOTS) {
1821             error_report("Specified number of memory slots %"
1822                          PRIu64" exceeds max supported %d",
1823                          machine->ram_slots, SPAPR_MAX_RAM_SLOTS);
1824             exit(1);
1825         }
1826 
1827         spapr->hotplug_memory.base = ROUND_UP(machine->ram_size,
1828                                               SPAPR_HOTPLUG_MEM_ALIGN);
1829         memory_region_init(&spapr->hotplug_memory.mr, OBJECT(spapr),
1830                            "hotplug-memory", hotplug_mem_size);
1831         memory_region_add_subregion(sysmem, spapr->hotplug_memory.base,
1832                                     &spapr->hotplug_memory.mr);
1833     }
1834 
1835     if (smc->dr_lmb_enabled) {
1836         spapr_create_lmb_dr_connectors(spapr);
1837     }
1838 
1839     filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, "spapr-rtas.bin");
1840     if (!filename) {
1841         error_report("Could not find LPAR rtas '%s'", "spapr-rtas.bin");
1842         exit(1);
1843     }
1844     spapr->rtas_size = get_image_size(filename);
1845     spapr->rtas_blob = g_malloc(spapr->rtas_size);
1846     if (load_image_size(filename, spapr->rtas_blob, spapr->rtas_size) < 0) {
1847         error_report("Could not load LPAR rtas '%s'", filename);
1848         exit(1);
1849     }
1850     if (spapr->rtas_size > RTAS_MAX_SIZE) {
1851         error_report("RTAS too big ! 0x%zx bytes (max is 0x%x)",
1852                      (size_t)spapr->rtas_size, RTAS_MAX_SIZE);
1853         exit(1);
1854     }
1855     g_free(filename);
1856 
1857     /* Set up EPOW events infrastructure */
1858     spapr_events_init(spapr);
1859 
1860     /* Set up the RTC RTAS interfaces */
1861     spapr_rtc_create(spapr);
1862 
1863     /* Set up VIO bus */
1864     spapr->vio_bus = spapr_vio_bus_init();
1865 
1866     for (i = 0; i < MAX_SERIAL_PORTS; i++) {
1867         if (serial_hds[i]) {
1868             spapr_vty_create(spapr->vio_bus, serial_hds[i]);
1869         }
1870     }
1871 
1872     /* We always have at least the nvram device on VIO */
1873     spapr_create_nvram(spapr);
1874 
1875     /* Set up PCI */
1876     spapr_pci_rtas_init();
1877 
1878     phb = spapr_create_phb(spapr, 0);
1879 
1880     for (i = 0; i < nb_nics; i++) {
1881         NICInfo *nd = &nd_table[i];
1882 
1883         if (!nd->model) {
1884             nd->model = g_strdup("ibmveth");
1885         }
1886 
1887         if (strcmp(nd->model, "ibmveth") == 0) {
1888             spapr_vlan_create(spapr->vio_bus, nd);
1889         } else {
1890             pci_nic_init_nofail(&nd_table[i], phb->bus, nd->model, NULL);
1891         }
1892     }
1893 
1894     for (i = 0; i <= drive_get_max_bus(IF_SCSI); i++) {
1895         spapr_vscsi_create(spapr->vio_bus);
1896     }
1897 
1898     /* Graphics */
1899     if (spapr_vga_init(phb->bus, &error_fatal)) {
1900         spapr->has_graphics = true;
1901         machine->usb |= defaults_enabled() && !machine->usb_disabled;
1902     }
1903 
1904     if (machine->usb) {
1905         if (smc->use_ohci_by_default) {
1906             pci_create_simple(phb->bus, -1, "pci-ohci");
1907         } else {
1908             pci_create_simple(phb->bus, -1, "nec-usb-xhci");
1909         }
1910 
1911         if (spapr->has_graphics) {
1912             USBBus *usb_bus = usb_bus_find(-1);
1913 
1914             usb_create_simple(usb_bus, "usb-kbd");
1915             usb_create_simple(usb_bus, "usb-mouse");
1916         }
1917     }
1918 
1919     if (spapr->rma_size < (MIN_RMA_SLOF << 20)) {
1920         error_report(
1921             "pSeries SLOF firmware requires >= %ldM guest RMA (Real Mode Area memory)",
1922             MIN_RMA_SLOF);
1923         exit(1);
1924     }
1925 
1926     if (kernel_filename) {
1927         uint64_t lowaddr = 0;
1928 
1929         kernel_size = load_elf(kernel_filename, translate_kernel_address, NULL,
1930                                NULL, &lowaddr, NULL, 1, PPC_ELF_MACHINE,
1931                                0, 0);
1932         if (kernel_size == ELF_LOAD_WRONG_ENDIAN) {
1933             kernel_size = load_elf(kernel_filename,
1934                                    translate_kernel_address, NULL,
1935                                    NULL, &lowaddr, NULL, 0, PPC_ELF_MACHINE,
1936                                    0, 0);
1937             kernel_le = kernel_size > 0;
1938         }
1939         if (kernel_size < 0) {
1940             error_report("error loading %s: %s",
1941                          kernel_filename, load_elf_strerror(kernel_size));
1942             exit(1);
1943         }
1944 
1945         /* load initrd */
1946         if (initrd_filename) {
1947             /* Try to locate the initrd in the gap between the kernel
1948              * and the firmware. Add a bit of space just in case
1949              */
1950             initrd_base = (KERNEL_LOAD_ADDR + kernel_size + 0x1ffff) & ~0xffff;
1951             initrd_size = load_image_targphys(initrd_filename, initrd_base,
1952                                               load_limit - initrd_base);
1953             if (initrd_size < 0) {
1954                 error_report("could not load initial ram disk '%s'",
1955                              initrd_filename);
1956                 exit(1);
1957             }
1958         } else {
1959             initrd_base = 0;
1960             initrd_size = 0;
1961         }
1962     }
1963 
1964     if (bios_name == NULL) {
1965         bios_name = FW_FILE_NAME;
1966     }
1967     filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, bios_name);
1968     if (!filename) {
1969         error_report("Could not find LPAR firmware '%s'", bios_name);
1970         exit(1);
1971     }
1972     fw_size = load_image_targphys(filename, 0, FW_MAX_SIZE);
1973     if (fw_size <= 0) {
1974         error_report("Could not load LPAR firmware '%s'", filename);
1975         exit(1);
1976     }
1977     g_free(filename);
1978 
1979     /* FIXME: Should register things through the MachineState's qdev
1980      * interface, this is a legacy from the sPAPREnvironment structure
1981      * which predated MachineState but had a similar function */
1982     vmstate_register(NULL, 0, &vmstate_spapr, spapr);
1983     register_savevm_live(NULL, "spapr/htab", -1, 1,
1984                          &savevm_htab_handlers, spapr);
1985 
1986     /* Prepare the device tree */
1987     spapr->fdt_skel = spapr_create_fdt_skel(initrd_base, initrd_size,
1988                                             kernel_size, kernel_le,
1989                                             kernel_cmdline,
1990                                             spapr->check_exception_irq);
1991     assert(spapr->fdt_skel != NULL);
1992 
1993     /* used by RTAS */
1994     QTAILQ_INIT(&spapr->ccs_list);
1995     qemu_register_reset(spapr_ccs_reset_hook, spapr);
1996 
1997     qemu_register_boot_set(spapr_boot_set, spapr);
1998 }
1999 
2000 static int spapr_kvm_type(const char *vm_type)
2001 {
2002     if (!vm_type) {
2003         return 0;
2004     }
2005 
2006     if (!strcmp(vm_type, "HV")) {
2007         return 1;
2008     }
2009 
2010     if (!strcmp(vm_type, "PR")) {
2011         return 2;
2012     }
2013 
2014     error_report("Unknown kvm-type specified '%s'", vm_type);
2015     exit(1);
2016 }
2017 
2018 /*
2019  * Implementation of an interface to adjust firmware path
2020  * for the bootindex property handling.
2021  */
2022 static char *spapr_get_fw_dev_path(FWPathProvider *p, BusState *bus,
2023                                    DeviceState *dev)
2024 {
2025 #define CAST(type, obj, name) \
2026     ((type *)object_dynamic_cast(OBJECT(obj), (name)))
2027     SCSIDevice *d = CAST(SCSIDevice,  dev, TYPE_SCSI_DEVICE);
2028     sPAPRPHBState *phb = CAST(sPAPRPHBState, dev, TYPE_SPAPR_PCI_HOST_BRIDGE);
2029 
2030     if (d) {
2031         void *spapr = CAST(void, bus->parent, "spapr-vscsi");
2032         VirtIOSCSI *virtio = CAST(VirtIOSCSI, bus->parent, TYPE_VIRTIO_SCSI);
2033         USBDevice *usb = CAST(USBDevice, bus->parent, TYPE_USB_DEVICE);
2034 
2035         if (spapr) {
2036             /*
2037              * Replace "channel@0/disk@0,0" with "disk@8000000000000000":
2038              * We use SRP luns of the form 8000 | (bus << 8) | (id << 5) | lun
2039              * in the top 16 bits of the 64-bit LUN
2040              */
2041             unsigned id = 0x8000 | (d->id << 8) | d->lun;
2042             return g_strdup_printf("%s@%"PRIX64, qdev_fw_name(dev),
2043                                    (uint64_t)id << 48);
2044         } else if (virtio) {
2045             /*
2046              * We use SRP luns of the form 01000000 | (target << 8) | lun
2047              * in the top 32 bits of the 64-bit LUN
2048              * Note: the quote above is from SLOF and it is wrong,
2049              * the actual binding is:
2050              * swap 0100 or 10 << or 20 << ( target lun-id -- srplun )
2051              */
2052             unsigned id = 0x1000000 | (d->id << 16) | d->lun;
2053             return g_strdup_printf("%s@%"PRIX64, qdev_fw_name(dev),
2054                                    (uint64_t)id << 32);
2055         } else if (usb) {
2056             /*
2057              * We use SRP luns of the form 01000000 | (usb-port << 16) | lun
2058              * in the top 32 bits of the 64-bit LUN
2059              */
2060             unsigned usb_port = atoi(usb->port->path);
2061             unsigned id = 0x1000000 | (usb_port << 16) | d->lun;
2062             return g_strdup_printf("%s@%"PRIX64, qdev_fw_name(dev),
2063                                    (uint64_t)id << 32);
2064         }
2065     }
2066 
2067     if (phb) {
2068         /* Replace "pci" with "pci@800000020000000" */
2069         return g_strdup_printf("pci@%"PRIX64, phb->buid);
2070     }
2071 
2072     return NULL;
2073 }
2074 
2075 static char *spapr_get_kvm_type(Object *obj, Error **errp)
2076 {
2077     sPAPRMachineState *spapr = SPAPR_MACHINE(obj);
2078 
2079     return g_strdup(spapr->kvm_type);
2080 }
2081 
2082 static void spapr_set_kvm_type(Object *obj, const char *value, Error **errp)
2083 {
2084     sPAPRMachineState *spapr = SPAPR_MACHINE(obj);
2085 
2086     g_free(spapr->kvm_type);
2087     spapr->kvm_type = g_strdup(value);
2088 }
2089 
2090 static void spapr_machine_initfn(Object *obj)
2091 {
2092     sPAPRMachineState *spapr = SPAPR_MACHINE(obj);
2093 
2094     spapr->htab_fd = -1;
2095     object_property_add_str(obj, "kvm-type",
2096                             spapr_get_kvm_type, spapr_set_kvm_type, NULL);
2097     object_property_set_description(obj, "kvm-type",
2098                                     "Specifies the KVM virtualization mode (HV, PR)",
2099                                     NULL);
2100 }
2101 
2102 static void spapr_machine_finalizefn(Object *obj)
2103 {
2104     sPAPRMachineState *spapr = SPAPR_MACHINE(obj);
2105 
2106     g_free(spapr->kvm_type);
2107 }
2108 
2109 static void ppc_cpu_do_nmi_on_cpu(void *arg)
2110 {
2111     CPUState *cs = arg;
2112 
2113     cpu_synchronize_state(cs);
2114     ppc_cpu_do_system_reset(cs);
2115 }
2116 
2117 static void spapr_nmi(NMIState *n, int cpu_index, Error **errp)
2118 {
2119     CPUState *cs;
2120 
2121     CPU_FOREACH(cs) {
2122         async_run_on_cpu(cs, ppc_cpu_do_nmi_on_cpu, cs);
2123     }
2124 }
2125 
2126 static void spapr_add_lmbs(DeviceState *dev, uint64_t addr, uint64_t size,
2127                            uint32_t node, Error **errp)
2128 {
2129     sPAPRDRConnector *drc;
2130     sPAPRDRConnectorClass *drck;
2131     uint32_t nr_lmbs = size/SPAPR_MEMORY_BLOCK_SIZE;
2132     int i, fdt_offset, fdt_size;
2133     void *fdt;
2134 
2135     /*
2136      * Check for DRC connectors and send hotplug notification to the
2137      * guest only in case of hotplugged memory. This allows cold plugged
2138      * memory to be specified at boot time.
2139      */
2140     if (!dev->hotplugged) {
2141         return;
2142     }
2143 
2144     for (i = 0; i < nr_lmbs; i++) {
2145         drc = spapr_dr_connector_by_id(SPAPR_DR_CONNECTOR_TYPE_LMB,
2146                 addr/SPAPR_MEMORY_BLOCK_SIZE);
2147         g_assert(drc);
2148 
2149         fdt = create_device_tree(&fdt_size);
2150         fdt_offset = spapr_populate_memory_node(fdt, node, addr,
2151                                                 SPAPR_MEMORY_BLOCK_SIZE);
2152 
2153         drck = SPAPR_DR_CONNECTOR_GET_CLASS(drc);
2154         drck->attach(drc, dev, fdt, fdt_offset, !dev->hotplugged, errp);
2155         addr += SPAPR_MEMORY_BLOCK_SIZE;
2156     }
2157     spapr_hotplug_req_add_by_count(SPAPR_DR_CONNECTOR_TYPE_LMB, nr_lmbs);
2158 }
2159 
2160 static void spapr_memory_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
2161                               uint32_t node, Error **errp)
2162 {
2163     Error *local_err = NULL;
2164     sPAPRMachineState *ms = SPAPR_MACHINE(hotplug_dev);
2165     PCDIMMDevice *dimm = PC_DIMM(dev);
2166     PCDIMMDeviceClass *ddc = PC_DIMM_GET_CLASS(dimm);
2167     MemoryRegion *mr = ddc->get_memory_region(dimm);
2168     uint64_t align = memory_region_get_alignment(mr);
2169     uint64_t size = memory_region_size(mr);
2170     uint64_t addr;
2171 
2172     if (size % SPAPR_MEMORY_BLOCK_SIZE) {
2173         error_setg(&local_err, "Hotplugged memory size must be a multiple of "
2174                       "%lld MB", SPAPR_MEMORY_BLOCK_SIZE/M_BYTE);
2175         goto out;
2176     }
2177 
2178     pc_dimm_memory_plug(dev, &ms->hotplug_memory, mr, align, &local_err);
2179     if (local_err) {
2180         goto out;
2181     }
2182 
2183     addr = object_property_get_int(OBJECT(dimm), PC_DIMM_ADDR_PROP, &local_err);
2184     if (local_err) {
2185         pc_dimm_memory_unplug(dev, &ms->hotplug_memory, mr);
2186         goto out;
2187     }
2188 
2189     spapr_add_lmbs(dev, addr, size, node, &error_abort);
2190 
2191 out:
2192     error_propagate(errp, local_err);
2193 }
2194 
2195 static void spapr_machine_device_plug(HotplugHandler *hotplug_dev,
2196                                       DeviceState *dev, Error **errp)
2197 {
2198     sPAPRMachineClass *smc = SPAPR_MACHINE_GET_CLASS(qdev_get_machine());
2199 
2200     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2201         int node;
2202 
2203         if (!smc->dr_lmb_enabled) {
2204             error_setg(errp, "Memory hotplug not supported for this machine");
2205             return;
2206         }
2207         node = object_property_get_int(OBJECT(dev), PC_DIMM_NODE_PROP, errp);
2208         if (*errp) {
2209             return;
2210         }
2211         if (node < 0 || node >= MAX_NODES) {
2212             error_setg(errp, "Invaild node %d", node);
2213             return;
2214         }
2215 
2216         /*
2217          * Currently PowerPC kernel doesn't allow hot-adding memory to
2218          * memory-less node, but instead will silently add the memory
2219          * to the first node that has some memory. This causes two
2220          * unexpected behaviours for the user.
2221          *
2222          * - Memory gets hotplugged to a different node than what the user
2223          *   specified.
2224          * - Since pc-dimm subsystem in QEMU still thinks that memory belongs
2225          *   to memory-less node, a reboot will set things accordingly
2226          *   and the previously hotplugged memory now ends in the right node.
2227          *   This appears as if some memory moved from one node to another.
2228          *
2229          * So until kernel starts supporting memory hotplug to memory-less
2230          * nodes, just prevent such attempts upfront in QEMU.
2231          */
2232         if (nb_numa_nodes && !numa_info[node].node_mem) {
2233             error_setg(errp, "Can't hotplug memory to memory-less node %d",
2234                        node);
2235             return;
2236         }
2237 
2238         spapr_memory_plug(hotplug_dev, dev, node, errp);
2239     }
2240 }
2241 
2242 static void spapr_machine_device_unplug(HotplugHandler *hotplug_dev,
2243                                       DeviceState *dev, Error **errp)
2244 {
2245     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2246         error_setg(errp, "Memory hot unplug not supported by sPAPR");
2247     }
2248 }
2249 
2250 static HotplugHandler *spapr_get_hotpug_handler(MachineState *machine,
2251                                              DeviceState *dev)
2252 {
2253     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2254         return HOTPLUG_HANDLER(machine);
2255     }
2256     return NULL;
2257 }
2258 
2259 static unsigned spapr_cpu_index_to_socket_id(unsigned cpu_index)
2260 {
2261     /* Allocate to NUMA nodes on a "socket" basis (not that concept of
2262      * socket means much for the paravirtualized PAPR platform) */
2263     return cpu_index / smp_threads / smp_cores;
2264 }
2265 
2266 static void spapr_machine_class_init(ObjectClass *oc, void *data)
2267 {
2268     MachineClass *mc = MACHINE_CLASS(oc);
2269     sPAPRMachineClass *smc = SPAPR_MACHINE_CLASS(oc);
2270     FWPathProviderClass *fwc = FW_PATH_PROVIDER_CLASS(oc);
2271     NMIClass *nc = NMI_CLASS(oc);
2272     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
2273 
2274     mc->desc = "pSeries Logical Partition (PAPR compliant)";
2275 
2276     /*
2277      * We set up the default / latest behaviour here.  The class_init
2278      * functions for the specific versioned machine types can override
2279      * these details for backwards compatibility
2280      */
2281     mc->init = ppc_spapr_init;
2282     mc->reset = ppc_spapr_reset;
2283     mc->block_default_type = IF_SCSI;
2284     mc->max_cpus = MAX_CPUMASK_BITS;
2285     mc->no_parallel = 1;
2286     mc->default_boot_order = "";
2287     mc->default_ram_size = 512 * M_BYTE;
2288     mc->kvm_type = spapr_kvm_type;
2289     mc->has_dynamic_sysbus = true;
2290     mc->pci_allow_0_address = true;
2291     mc->get_hotplug_handler = spapr_get_hotpug_handler;
2292     hc->plug = spapr_machine_device_plug;
2293     hc->unplug = spapr_machine_device_unplug;
2294     mc->cpu_index_to_socket_id = spapr_cpu_index_to_socket_id;
2295 
2296     smc->dr_lmb_enabled = true;
2297     fwc->get_dev_path = spapr_get_fw_dev_path;
2298     nc->nmi_monitor_handler = spapr_nmi;
2299 }
2300 
2301 static const TypeInfo spapr_machine_info = {
2302     .name          = TYPE_SPAPR_MACHINE,
2303     .parent        = TYPE_MACHINE,
2304     .abstract      = true,
2305     .instance_size = sizeof(sPAPRMachineState),
2306     .instance_init = spapr_machine_initfn,
2307     .instance_finalize = spapr_machine_finalizefn,
2308     .class_size    = sizeof(sPAPRMachineClass),
2309     .class_init    = spapr_machine_class_init,
2310     .interfaces = (InterfaceInfo[]) {
2311         { TYPE_FW_PATH_PROVIDER },
2312         { TYPE_NMI },
2313         { TYPE_HOTPLUG_HANDLER },
2314         { }
2315     },
2316 };
2317 
2318 #define DEFINE_SPAPR_MACHINE(suffix, verstr, latest)                 \
2319     static void spapr_machine_##suffix##_class_init(ObjectClass *oc, \
2320                                                     void *data)      \
2321     {                                                                \
2322         MachineClass *mc = MACHINE_CLASS(oc);                        \
2323         spapr_machine_##suffix##_class_options(mc);                  \
2324         if (latest) {                                                \
2325             mc->alias = "pseries";                                   \
2326             mc->is_default = 1;                                      \
2327         }                                                            \
2328     }                                                                \
2329     static void spapr_machine_##suffix##_instance_init(Object *obj)  \
2330     {                                                                \
2331         MachineState *machine = MACHINE(obj);                        \
2332         spapr_machine_##suffix##_instance_options(machine);          \
2333     }                                                                \
2334     static const TypeInfo spapr_machine_##suffix##_info = {          \
2335         .name = MACHINE_TYPE_NAME("pseries-" verstr),                \
2336         .parent = TYPE_SPAPR_MACHINE,                                \
2337         .class_init = spapr_machine_##suffix##_class_init,           \
2338         .instance_init = spapr_machine_##suffix##_instance_init,     \
2339     };                                                               \
2340     static void spapr_machine_register_##suffix(void)                \
2341     {                                                                \
2342         type_register(&spapr_machine_##suffix##_info);               \
2343     }                                                                \
2344     type_init(spapr_machine_register_##suffix)
2345 
2346 /*
2347  * pseries-2.6
2348  */
2349 static void spapr_machine_2_6_instance_options(MachineState *machine)
2350 {
2351 }
2352 
2353 static void spapr_machine_2_6_class_options(MachineClass *mc)
2354 {
2355     /* Defaults for the latest behaviour inherited from the base class */
2356 }
2357 
2358 DEFINE_SPAPR_MACHINE(2_6, "2.6", true);
2359 
2360 /*
2361  * pseries-2.5
2362  */
2363 #define SPAPR_COMPAT_2_5 \
2364     HW_COMPAT_2_5 \
2365     { \
2366         .driver   = "spapr-vlan", \
2367         .property = "use-rx-buffer-pools", \
2368         .value    = "off", \
2369     },
2370 
2371 static void spapr_machine_2_5_instance_options(MachineState *machine)
2372 {
2373 }
2374 
2375 static void spapr_machine_2_5_class_options(MachineClass *mc)
2376 {
2377     sPAPRMachineClass *smc = SPAPR_MACHINE_CLASS(mc);
2378 
2379     spapr_machine_2_6_class_options(mc);
2380     smc->use_ohci_by_default = true;
2381     SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_5);
2382 }
2383 
2384 DEFINE_SPAPR_MACHINE(2_5, "2.5", false);
2385 
2386 /*
2387  * pseries-2.4
2388  */
2389 #define SPAPR_COMPAT_2_4 \
2390         SPAPR_COMPAT_2_5 \
2391         HW_COMPAT_2_4
2392 
2393 static void spapr_machine_2_4_instance_options(MachineState *machine)
2394 {
2395     spapr_machine_2_5_instance_options(machine);
2396 }
2397 
2398 static void spapr_machine_2_4_class_options(MachineClass *mc)
2399 {
2400     sPAPRMachineClass *smc = SPAPR_MACHINE_CLASS(mc);
2401 
2402     spapr_machine_2_5_class_options(mc);
2403     smc->dr_lmb_enabled = false;
2404     SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_4);
2405 }
2406 
2407 DEFINE_SPAPR_MACHINE(2_4, "2.4", false);
2408 
2409 /*
2410  * pseries-2.3
2411  */
2412 #define SPAPR_COMPAT_2_3 \
2413         SPAPR_COMPAT_2_4 \
2414         HW_COMPAT_2_3 \
2415         {\
2416             .driver   = "spapr-pci-host-bridge",\
2417             .property = "dynamic-reconfiguration",\
2418             .value    = "off",\
2419         },
2420 
2421 static void spapr_machine_2_3_instance_options(MachineState *machine)
2422 {
2423     spapr_machine_2_4_instance_options(machine);
2424     savevm_skip_section_footers();
2425     global_state_set_optional();
2426     savevm_skip_configuration();
2427 }
2428 
2429 static void spapr_machine_2_3_class_options(MachineClass *mc)
2430 {
2431     spapr_machine_2_4_class_options(mc);
2432     SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_3);
2433 }
2434 DEFINE_SPAPR_MACHINE(2_3, "2.3", false);
2435 
2436 /*
2437  * pseries-2.2
2438  */
2439 
2440 #define SPAPR_COMPAT_2_2 \
2441         SPAPR_COMPAT_2_3 \
2442         HW_COMPAT_2_2 \
2443         {\
2444             .driver   = TYPE_SPAPR_PCI_HOST_BRIDGE,\
2445             .property = "mem_win_size",\
2446             .value    = "0x20000000",\
2447         },
2448 
2449 static void spapr_machine_2_2_instance_options(MachineState *machine)
2450 {
2451     spapr_machine_2_3_instance_options(machine);
2452     machine->suppress_vmdesc = true;
2453 }
2454 
2455 static void spapr_machine_2_2_class_options(MachineClass *mc)
2456 {
2457     spapr_machine_2_3_class_options(mc);
2458     SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_2);
2459 }
2460 DEFINE_SPAPR_MACHINE(2_2, "2.2", false);
2461 
2462 /*
2463  * pseries-2.1
2464  */
2465 #define SPAPR_COMPAT_2_1 \
2466         SPAPR_COMPAT_2_2 \
2467         HW_COMPAT_2_1
2468 
2469 static void spapr_machine_2_1_instance_options(MachineState *machine)
2470 {
2471     spapr_machine_2_2_instance_options(machine);
2472 }
2473 
2474 static void spapr_machine_2_1_class_options(MachineClass *mc)
2475 {
2476     spapr_machine_2_2_class_options(mc);
2477     SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_1);
2478 }
2479 DEFINE_SPAPR_MACHINE(2_1, "2.1", false);
2480 
2481 static void spapr_machine_register_types(void)
2482 {
2483     type_register_static(&spapr_machine_info);
2484 }
2485 
2486 type_init(spapr_machine_register_types)
2487