xref: /qemu/hw/ppc/spapr_hcall.c (revision 5db05230)
1 #include "qemu/osdep.h"
2 #include "qemu/cutils.h"
3 #include "qapi/error.h"
4 #include "sysemu/hw_accel.h"
5 #include "sysemu/runstate.h"
6 #include "sysemu/tcg.h"
7 #include "qemu/log.h"
8 #include "qemu/main-loop.h"
9 #include "qemu/module.h"
10 #include "qemu/error-report.h"
11 #include "exec/tb-flush.h"
12 #include "helper_regs.h"
13 #include "hw/ppc/ppc.h"
14 #include "hw/ppc/spapr.h"
15 #include "hw/ppc/spapr_cpu_core.h"
16 #include "hw/ppc/spapr_nested.h"
17 #include "mmu-hash64.h"
18 #include "cpu-models.h"
19 #include "trace.h"
20 #include "kvm_ppc.h"
21 #include "hw/ppc/fdt.h"
22 #include "hw/ppc/spapr_ovec.h"
23 #include "hw/ppc/spapr_numa.h"
24 #include "mmu-book3s-v3.h"
25 #include "hw/mem/memory-device.h"
26 
27 bool is_ram_address(SpaprMachineState *spapr, hwaddr addr)
28 {
29     MachineState *machine = MACHINE(spapr);
30     DeviceMemoryState *dms = machine->device_memory;
31 
32     if (addr < machine->ram_size) {
33         return true;
34     }
35     if (dms && (addr >= dms->base)
36         && ((addr - dms->base) < memory_region_size(&dms->mr))) {
37         return true;
38     }
39 
40     return false;
41 }
42 
43 /* Convert a return code from the KVM ioctl()s implementing resize HPT
44  * into a PAPR hypercall return code */
45 static target_ulong resize_hpt_convert_rc(int ret)
46 {
47     if (ret >= 100000) {
48         return H_LONG_BUSY_ORDER_100_SEC;
49     } else if (ret >= 10000) {
50         return H_LONG_BUSY_ORDER_10_SEC;
51     } else if (ret >= 1000) {
52         return H_LONG_BUSY_ORDER_1_SEC;
53     } else if (ret >= 100) {
54         return H_LONG_BUSY_ORDER_100_MSEC;
55     } else if (ret >= 10) {
56         return H_LONG_BUSY_ORDER_10_MSEC;
57     } else if (ret > 0) {
58         return H_LONG_BUSY_ORDER_1_MSEC;
59     }
60 
61     switch (ret) {
62     case 0:
63         return H_SUCCESS;
64     case -EPERM:
65         return H_AUTHORITY;
66     case -EINVAL:
67         return H_PARAMETER;
68     case -ENXIO:
69         return H_CLOSED;
70     case -ENOSPC:
71         return H_PTEG_FULL;
72     case -EBUSY:
73         return H_BUSY;
74     case -ENOMEM:
75         return H_NO_MEM;
76     default:
77         return H_HARDWARE;
78     }
79 }
80 
81 static target_ulong h_resize_hpt_prepare(PowerPCCPU *cpu,
82                                          SpaprMachineState *spapr,
83                                          target_ulong opcode,
84                                          target_ulong *args)
85 {
86     target_ulong flags = args[0];
87     int shift = args[1];
88     uint64_t current_ram_size;
89     int rc;
90 
91     if (spapr->resize_hpt == SPAPR_RESIZE_HPT_DISABLED) {
92         return H_AUTHORITY;
93     }
94 
95     if (!spapr->htab_shift) {
96         /* Radix guest, no HPT */
97         return H_NOT_AVAILABLE;
98     }
99 
100     trace_spapr_h_resize_hpt_prepare(flags, shift);
101 
102     if (flags != 0) {
103         return H_PARAMETER;
104     }
105 
106     if (shift && ((shift < 18) || (shift > 46))) {
107         return H_PARAMETER;
108     }
109 
110     current_ram_size = MACHINE(spapr)->ram_size + get_plugged_memory_size();
111 
112     /* We only allow the guest to allocate an HPT one order above what
113      * we'd normally give them (to stop a small guest claiming a huge
114      * chunk of resources in the HPT */
115     if (shift > (spapr_hpt_shift_for_ramsize(current_ram_size) + 1)) {
116         return H_RESOURCE;
117     }
118 
119     rc = kvmppc_resize_hpt_prepare(cpu, flags, shift);
120     if (rc != -ENOSYS) {
121         return resize_hpt_convert_rc(rc);
122     }
123 
124     if (kvm_enabled()) {
125         return H_HARDWARE;
126     }
127 
128     return softmmu_resize_hpt_prepare(cpu, spapr, shift);
129 }
130 
131 static void do_push_sregs_to_kvm_pr(CPUState *cs, run_on_cpu_data data)
132 {
133     int ret;
134 
135     cpu_synchronize_state(cs);
136 
137     ret = kvmppc_put_books_sregs(POWERPC_CPU(cs));
138     if (ret < 0) {
139         error_report("failed to push sregs to KVM: %s", strerror(-ret));
140         exit(1);
141     }
142 }
143 
144 void push_sregs_to_kvm_pr(SpaprMachineState *spapr)
145 {
146     CPUState *cs;
147 
148     /*
149      * This is a hack for the benefit of KVM PR - it abuses the SDR1
150      * slot in kvm_sregs to communicate the userspace address of the
151      * HPT
152      */
153     if (!kvm_enabled() || !spapr->htab) {
154         return;
155     }
156 
157     CPU_FOREACH(cs) {
158         run_on_cpu(cs, do_push_sregs_to_kvm_pr, RUN_ON_CPU_NULL);
159     }
160 }
161 
162 static target_ulong h_resize_hpt_commit(PowerPCCPU *cpu,
163                                         SpaprMachineState *spapr,
164                                         target_ulong opcode,
165                                         target_ulong *args)
166 {
167     target_ulong flags = args[0];
168     target_ulong shift = args[1];
169     int rc;
170 
171     if (spapr->resize_hpt == SPAPR_RESIZE_HPT_DISABLED) {
172         return H_AUTHORITY;
173     }
174 
175     if (!spapr->htab_shift) {
176         /* Radix guest, no HPT */
177         return H_NOT_AVAILABLE;
178     }
179 
180     trace_spapr_h_resize_hpt_commit(flags, shift);
181 
182     rc = kvmppc_resize_hpt_commit(cpu, flags, shift);
183     if (rc != -ENOSYS) {
184         rc = resize_hpt_convert_rc(rc);
185         if (rc == H_SUCCESS) {
186             /* Need to set the new htab_shift in the machine state */
187             spapr->htab_shift = shift;
188         }
189         return rc;
190     }
191 
192     if (kvm_enabled()) {
193         return H_HARDWARE;
194     }
195 
196     return softmmu_resize_hpt_commit(cpu, spapr, flags, shift);
197 }
198 
199 
200 
201 static target_ulong h_set_sprg0(PowerPCCPU *cpu, SpaprMachineState *spapr,
202                                 target_ulong opcode, target_ulong *args)
203 {
204     cpu_synchronize_state(CPU(cpu));
205     cpu->env.spr[SPR_SPRG0] = args[0];
206 
207     return H_SUCCESS;
208 }
209 
210 static target_ulong h_set_dabr(PowerPCCPU *cpu, SpaprMachineState *spapr,
211                                target_ulong opcode, target_ulong *args)
212 {
213     if (!ppc_has_spr(cpu, SPR_DABR)) {
214         return H_HARDWARE;              /* DABR register not available */
215     }
216     cpu_synchronize_state(CPU(cpu));
217 
218     if (ppc_has_spr(cpu, SPR_DABRX)) {
219         cpu->env.spr[SPR_DABRX] = 0x3;  /* Use Problem and Privileged state */
220     } else if (!(args[0] & 0x4)) {      /* Breakpoint Translation set? */
221         return H_RESERVED_DABR;
222     }
223 
224     cpu->env.spr[SPR_DABR] = args[0];
225     return H_SUCCESS;
226 }
227 
228 static target_ulong h_set_xdabr(PowerPCCPU *cpu, SpaprMachineState *spapr,
229                                 target_ulong opcode, target_ulong *args)
230 {
231     target_ulong dabrx = args[1];
232 
233     if (!ppc_has_spr(cpu, SPR_DABR) || !ppc_has_spr(cpu, SPR_DABRX)) {
234         return H_HARDWARE;
235     }
236 
237     if ((dabrx & ~0xfULL) != 0 || (dabrx & H_DABRX_HYPERVISOR) != 0
238         || (dabrx & (H_DABRX_KERNEL | H_DABRX_USER)) == 0) {
239         return H_PARAMETER;
240     }
241 
242     cpu_synchronize_state(CPU(cpu));
243     cpu->env.spr[SPR_DABRX] = dabrx;
244     cpu->env.spr[SPR_DABR] = args[0];
245 
246     return H_SUCCESS;
247 }
248 
249 static target_ulong h_page_init(PowerPCCPU *cpu, SpaprMachineState *spapr,
250                                 target_ulong opcode, target_ulong *args)
251 {
252     target_ulong flags = args[0];
253     hwaddr dst = args[1];
254     hwaddr src = args[2];
255     hwaddr len = TARGET_PAGE_SIZE;
256     uint8_t *pdst, *psrc;
257     target_long ret = H_SUCCESS;
258 
259     if (flags & ~(H_ICACHE_SYNCHRONIZE | H_ICACHE_INVALIDATE
260                   | H_COPY_PAGE | H_ZERO_PAGE)) {
261         qemu_log_mask(LOG_UNIMP, "h_page_init: Bad flags (" TARGET_FMT_lx "\n",
262                       flags);
263         return H_PARAMETER;
264     }
265 
266     /* Map-in destination */
267     if (!is_ram_address(spapr, dst) || (dst & ~TARGET_PAGE_MASK) != 0) {
268         return H_PARAMETER;
269     }
270     pdst = cpu_physical_memory_map(dst, &len, true);
271     if (!pdst || len != TARGET_PAGE_SIZE) {
272         return H_PARAMETER;
273     }
274 
275     if (flags & H_COPY_PAGE) {
276         /* Map-in source, copy to destination, and unmap source again */
277         if (!is_ram_address(spapr, src) || (src & ~TARGET_PAGE_MASK) != 0) {
278             ret = H_PARAMETER;
279             goto unmap_out;
280         }
281         psrc = cpu_physical_memory_map(src, &len, false);
282         if (!psrc || len != TARGET_PAGE_SIZE) {
283             ret = H_PARAMETER;
284             goto unmap_out;
285         }
286         memcpy(pdst, psrc, len);
287         cpu_physical_memory_unmap(psrc, len, 0, len);
288     } else if (flags & H_ZERO_PAGE) {
289         memset(pdst, 0, len);          /* Just clear the destination page */
290     }
291 
292     if (kvm_enabled() && (flags & H_ICACHE_SYNCHRONIZE) != 0) {
293         kvmppc_dcbst_range(cpu, pdst, len);
294     }
295     if (flags & (H_ICACHE_SYNCHRONIZE | H_ICACHE_INVALIDATE)) {
296         if (kvm_enabled()) {
297             kvmppc_icbi_range(cpu, pdst, len);
298         } else {
299             tb_flush(CPU(cpu));
300         }
301     }
302 
303 unmap_out:
304     cpu_physical_memory_unmap(pdst, TARGET_PAGE_SIZE, 1, len);
305     return ret;
306 }
307 
308 #define FLAGS_REGISTER_VPA         0x0000200000000000ULL
309 #define FLAGS_REGISTER_DTL         0x0000400000000000ULL
310 #define FLAGS_REGISTER_SLBSHADOW   0x0000600000000000ULL
311 #define FLAGS_DEREGISTER_VPA       0x0000a00000000000ULL
312 #define FLAGS_DEREGISTER_DTL       0x0000c00000000000ULL
313 #define FLAGS_DEREGISTER_SLBSHADOW 0x0000e00000000000ULL
314 
315 static target_ulong register_vpa(PowerPCCPU *cpu, target_ulong vpa)
316 {
317     CPUState *cs = CPU(cpu);
318     CPUPPCState *env = &cpu->env;
319     SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
320     uint16_t size;
321     uint8_t tmp;
322 
323     if (vpa == 0) {
324         hcall_dprintf("Can't cope with registering a VPA at logical 0\n");
325         return H_HARDWARE;
326     }
327 
328     if (vpa % env->dcache_line_size) {
329         return H_PARAMETER;
330     }
331     /* FIXME: bounds check the address */
332 
333     size = lduw_be_phys(cs->as, vpa + 0x4);
334 
335     if (size < VPA_MIN_SIZE) {
336         return H_PARAMETER;
337     }
338 
339     /* VPA is not allowed to cross a page boundary */
340     if ((vpa / 4096) != ((vpa + size - 1) / 4096)) {
341         return H_PARAMETER;
342     }
343 
344     spapr_cpu->vpa_addr = vpa;
345 
346     tmp = ldub_phys(cs->as, spapr_cpu->vpa_addr + VPA_SHARED_PROC_OFFSET);
347     tmp |= VPA_SHARED_PROC_VAL;
348     stb_phys(cs->as, spapr_cpu->vpa_addr + VPA_SHARED_PROC_OFFSET, tmp);
349 
350     return H_SUCCESS;
351 }
352 
353 static target_ulong deregister_vpa(PowerPCCPU *cpu, target_ulong vpa)
354 {
355     SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
356 
357     if (spapr_cpu->slb_shadow_addr) {
358         return H_RESOURCE;
359     }
360 
361     if (spapr_cpu->dtl_addr) {
362         return H_RESOURCE;
363     }
364 
365     spapr_cpu->vpa_addr = 0;
366     return H_SUCCESS;
367 }
368 
369 static target_ulong register_slb_shadow(PowerPCCPU *cpu, target_ulong addr)
370 {
371     SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
372     uint32_t size;
373 
374     if (addr == 0) {
375         hcall_dprintf("Can't cope with SLB shadow at logical 0\n");
376         return H_HARDWARE;
377     }
378 
379     size = ldl_be_phys(CPU(cpu)->as, addr + 0x4);
380     if (size < 0x8) {
381         return H_PARAMETER;
382     }
383 
384     if ((addr / 4096) != ((addr + size - 1) / 4096)) {
385         return H_PARAMETER;
386     }
387 
388     if (!spapr_cpu->vpa_addr) {
389         return H_RESOURCE;
390     }
391 
392     spapr_cpu->slb_shadow_addr = addr;
393     spapr_cpu->slb_shadow_size = size;
394 
395     return H_SUCCESS;
396 }
397 
398 static target_ulong deregister_slb_shadow(PowerPCCPU *cpu, target_ulong addr)
399 {
400     SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
401 
402     spapr_cpu->slb_shadow_addr = 0;
403     spapr_cpu->slb_shadow_size = 0;
404     return H_SUCCESS;
405 }
406 
407 static target_ulong register_dtl(PowerPCCPU *cpu, target_ulong addr)
408 {
409     SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
410     uint32_t size;
411 
412     if (addr == 0) {
413         hcall_dprintf("Can't cope with DTL at logical 0\n");
414         return H_HARDWARE;
415     }
416 
417     size = ldl_be_phys(CPU(cpu)->as, addr + 0x4);
418 
419     if (size < 48) {
420         return H_PARAMETER;
421     }
422 
423     if (!spapr_cpu->vpa_addr) {
424         return H_RESOURCE;
425     }
426 
427     spapr_cpu->dtl_addr = addr;
428     spapr_cpu->dtl_size = size;
429 
430     return H_SUCCESS;
431 }
432 
433 static target_ulong deregister_dtl(PowerPCCPU *cpu, target_ulong addr)
434 {
435     SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
436 
437     spapr_cpu->dtl_addr = 0;
438     spapr_cpu->dtl_size = 0;
439 
440     return H_SUCCESS;
441 }
442 
443 static target_ulong h_register_vpa(PowerPCCPU *cpu, SpaprMachineState *spapr,
444                                    target_ulong opcode, target_ulong *args)
445 {
446     target_ulong flags = args[0];
447     target_ulong procno = args[1];
448     target_ulong vpa = args[2];
449     target_ulong ret = H_PARAMETER;
450     PowerPCCPU *tcpu;
451 
452     tcpu = spapr_find_cpu(procno);
453     if (!tcpu) {
454         return H_PARAMETER;
455     }
456 
457     switch (flags) {
458     case FLAGS_REGISTER_VPA:
459         ret = register_vpa(tcpu, vpa);
460         break;
461 
462     case FLAGS_DEREGISTER_VPA:
463         ret = deregister_vpa(tcpu, vpa);
464         break;
465 
466     case FLAGS_REGISTER_SLBSHADOW:
467         ret = register_slb_shadow(tcpu, vpa);
468         break;
469 
470     case FLAGS_DEREGISTER_SLBSHADOW:
471         ret = deregister_slb_shadow(tcpu, vpa);
472         break;
473 
474     case FLAGS_REGISTER_DTL:
475         ret = register_dtl(tcpu, vpa);
476         break;
477 
478     case FLAGS_DEREGISTER_DTL:
479         ret = deregister_dtl(tcpu, vpa);
480         break;
481     }
482 
483     return ret;
484 }
485 
486 static target_ulong h_cede(PowerPCCPU *cpu, SpaprMachineState *spapr,
487                            target_ulong opcode, target_ulong *args)
488 {
489     CPUPPCState *env = &cpu->env;
490     CPUState *cs = CPU(cpu);
491     SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
492 
493     env->msr |= (1ULL << MSR_EE);
494     hreg_compute_hflags(env);
495     ppc_maybe_interrupt(env);
496 
497     if (spapr_cpu->prod) {
498         spapr_cpu->prod = false;
499         return H_SUCCESS;
500     }
501 
502     if (!cpu_has_work(cs)) {
503         cs->halted = 1;
504         cs->exception_index = EXCP_HLT;
505         cs->exit_request = 1;
506         ppc_maybe_interrupt(env);
507     }
508 
509     return H_SUCCESS;
510 }
511 
512 /*
513  * Confer to self, aka join. Cede could use the same pattern as well, if
514  * EXCP_HLT can be changed to ECXP_HALTED.
515  */
516 static target_ulong h_confer_self(PowerPCCPU *cpu)
517 {
518     CPUState *cs = CPU(cpu);
519     SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
520 
521     if (spapr_cpu->prod) {
522         spapr_cpu->prod = false;
523         return H_SUCCESS;
524     }
525     cs->halted = 1;
526     cs->exception_index = EXCP_HALTED;
527     cs->exit_request = 1;
528     ppc_maybe_interrupt(&cpu->env);
529 
530     return H_SUCCESS;
531 }
532 
533 static target_ulong h_join(PowerPCCPU *cpu, SpaprMachineState *spapr,
534                            target_ulong opcode, target_ulong *args)
535 {
536     CPUPPCState *env = &cpu->env;
537     CPUState *cs;
538     bool last_unjoined = true;
539 
540     if (env->msr & (1ULL << MSR_EE)) {
541         return H_BAD_MODE;
542     }
543 
544     /*
545      * Must not join the last CPU running. Interestingly, no such restriction
546      * for H_CONFER-to-self, but that is probably not intended to be used
547      * when H_JOIN is available.
548      */
549     CPU_FOREACH(cs) {
550         PowerPCCPU *c = POWERPC_CPU(cs);
551         CPUPPCState *e = &c->env;
552         if (c == cpu) {
553             continue;
554         }
555 
556         /* Don't have a way to indicate joined, so use halted && MSR[EE]=0 */
557         if (!cs->halted || (e->msr & (1ULL << MSR_EE))) {
558             last_unjoined = false;
559             break;
560         }
561     }
562     if (last_unjoined) {
563         return H_CONTINUE;
564     }
565 
566     return h_confer_self(cpu);
567 }
568 
569 static target_ulong h_confer(PowerPCCPU *cpu, SpaprMachineState *spapr,
570                            target_ulong opcode, target_ulong *args)
571 {
572     target_long target = args[0];
573     uint32_t dispatch = args[1];
574     CPUState *cs = CPU(cpu);
575     SpaprCpuState *spapr_cpu;
576 
577     /*
578      * -1 means confer to all other CPUs without dispatch counter check,
579      *  otherwise it's a targeted confer.
580      */
581     if (target != -1) {
582         PowerPCCPU *target_cpu = spapr_find_cpu(target);
583         uint32_t target_dispatch;
584 
585         if (!target_cpu) {
586             return H_PARAMETER;
587         }
588 
589         /*
590          * target == self is a special case, we wait until prodded, without
591          * dispatch counter check.
592          */
593         if (cpu == target_cpu) {
594             return h_confer_self(cpu);
595         }
596 
597         spapr_cpu = spapr_cpu_state(target_cpu);
598         if (!spapr_cpu->vpa_addr || ((dispatch & 1) == 0)) {
599             return H_SUCCESS;
600         }
601 
602         target_dispatch = ldl_be_phys(cs->as,
603                                   spapr_cpu->vpa_addr + VPA_DISPATCH_COUNTER);
604         if (target_dispatch != dispatch) {
605             return H_SUCCESS;
606         }
607 
608         /*
609          * The targeted confer does not do anything special beyond yielding
610          * the current vCPU, but even this should be better than nothing.
611          * At least for single-threaded tcg, it gives the target a chance to
612          * run before we run again. Multi-threaded tcg does not really do
613          * anything with EXCP_YIELD yet.
614          */
615     }
616 
617     cs->exception_index = EXCP_YIELD;
618     cs->exit_request = 1;
619     cpu_loop_exit(cs);
620 
621     return H_SUCCESS;
622 }
623 
624 static target_ulong h_prod(PowerPCCPU *cpu, SpaprMachineState *spapr,
625                            target_ulong opcode, target_ulong *args)
626 {
627     target_long target = args[0];
628     PowerPCCPU *tcpu;
629     CPUState *cs;
630     SpaprCpuState *spapr_cpu;
631 
632     tcpu = spapr_find_cpu(target);
633     cs = CPU(tcpu);
634     if (!cs) {
635         return H_PARAMETER;
636     }
637 
638     spapr_cpu = spapr_cpu_state(tcpu);
639     spapr_cpu->prod = true;
640     cs->halted = 0;
641     ppc_maybe_interrupt(&cpu->env);
642     qemu_cpu_kick(cs);
643 
644     return H_SUCCESS;
645 }
646 
647 static target_ulong h_rtas(PowerPCCPU *cpu, SpaprMachineState *spapr,
648                            target_ulong opcode, target_ulong *args)
649 {
650     target_ulong rtas_r3 = args[0];
651     uint32_t token = rtas_ld(rtas_r3, 0);
652     uint32_t nargs = rtas_ld(rtas_r3, 1);
653     uint32_t nret = rtas_ld(rtas_r3, 2);
654 
655     return spapr_rtas_call(cpu, spapr, token, nargs, rtas_r3 + 12,
656                            nret, rtas_r3 + 12 + 4*nargs);
657 }
658 
659 static target_ulong h_logical_load(PowerPCCPU *cpu, SpaprMachineState *spapr,
660                                    target_ulong opcode, target_ulong *args)
661 {
662     CPUState *cs = CPU(cpu);
663     target_ulong size = args[0];
664     target_ulong addr = args[1];
665 
666     switch (size) {
667     case 1:
668         args[0] = ldub_phys(cs->as, addr);
669         return H_SUCCESS;
670     case 2:
671         args[0] = lduw_phys(cs->as, addr);
672         return H_SUCCESS;
673     case 4:
674         args[0] = ldl_phys(cs->as, addr);
675         return H_SUCCESS;
676     case 8:
677         args[0] = ldq_phys(cs->as, addr);
678         return H_SUCCESS;
679     }
680     return H_PARAMETER;
681 }
682 
683 static target_ulong h_logical_store(PowerPCCPU *cpu, SpaprMachineState *spapr,
684                                     target_ulong opcode, target_ulong *args)
685 {
686     CPUState *cs = CPU(cpu);
687 
688     target_ulong size = args[0];
689     target_ulong addr = args[1];
690     target_ulong val  = args[2];
691 
692     switch (size) {
693     case 1:
694         stb_phys(cs->as, addr, val);
695         return H_SUCCESS;
696     case 2:
697         stw_phys(cs->as, addr, val);
698         return H_SUCCESS;
699     case 4:
700         stl_phys(cs->as, addr, val);
701         return H_SUCCESS;
702     case 8:
703         stq_phys(cs->as, addr, val);
704         return H_SUCCESS;
705     }
706     return H_PARAMETER;
707 }
708 
709 static target_ulong h_logical_memop(PowerPCCPU *cpu, SpaprMachineState *spapr,
710                                     target_ulong opcode, target_ulong *args)
711 {
712     CPUState *cs = CPU(cpu);
713 
714     target_ulong dst   = args[0]; /* Destination address */
715     target_ulong src   = args[1]; /* Source address */
716     target_ulong esize = args[2]; /* Element size (0=1,1=2,2=4,3=8) */
717     target_ulong count = args[3]; /* Element count */
718     target_ulong op    = args[4]; /* 0 = copy, 1 = invert */
719     uint64_t tmp;
720     unsigned int mask = (1 << esize) - 1;
721     int step = 1 << esize;
722 
723     if (count > 0x80000000) {
724         return H_PARAMETER;
725     }
726 
727     if ((dst & mask) || (src & mask) || (op > 1)) {
728         return H_PARAMETER;
729     }
730 
731     if (dst >= src && dst < (src + (count << esize))) {
732             dst = dst + ((count - 1) << esize);
733             src = src + ((count - 1) << esize);
734             step = -step;
735     }
736 
737     while (count--) {
738         switch (esize) {
739         case 0:
740             tmp = ldub_phys(cs->as, src);
741             break;
742         case 1:
743             tmp = lduw_phys(cs->as, src);
744             break;
745         case 2:
746             tmp = ldl_phys(cs->as, src);
747             break;
748         case 3:
749             tmp = ldq_phys(cs->as, src);
750             break;
751         default:
752             return H_PARAMETER;
753         }
754         if (op == 1) {
755             tmp = ~tmp;
756         }
757         switch (esize) {
758         case 0:
759             stb_phys(cs->as, dst, tmp);
760             break;
761         case 1:
762             stw_phys(cs->as, dst, tmp);
763             break;
764         case 2:
765             stl_phys(cs->as, dst, tmp);
766             break;
767         case 3:
768             stq_phys(cs->as, dst, tmp);
769             break;
770         }
771         dst = dst + step;
772         src = src + step;
773     }
774 
775     return H_SUCCESS;
776 }
777 
778 static target_ulong h_logical_icbi(PowerPCCPU *cpu, SpaprMachineState *spapr,
779                                    target_ulong opcode, target_ulong *args)
780 {
781     /* Nothing to do on emulation, KVM will trap this in the kernel */
782     return H_SUCCESS;
783 }
784 
785 static target_ulong h_logical_dcbf(PowerPCCPU *cpu, SpaprMachineState *spapr,
786                                    target_ulong opcode, target_ulong *args)
787 {
788     /* Nothing to do on emulation, KVM will trap this in the kernel */
789     return H_SUCCESS;
790 }
791 
792 static target_ulong h_set_mode_resource_set_ciabr(PowerPCCPU *cpu,
793                                                   SpaprMachineState *spapr,
794                                                   target_ulong mflags,
795                                                   target_ulong value1,
796                                                   target_ulong value2)
797 {
798     CPUPPCState *env = &cpu->env;
799 
800     assert(tcg_enabled()); /* KVM will have handled this */
801 
802     if (mflags) {
803         return H_UNSUPPORTED_FLAG;
804     }
805     if (value2) {
806         return H_P4;
807     }
808     if ((value1 & PPC_BITMASK(62, 63)) == 0x3) {
809         return H_P3;
810     }
811 
812     ppc_store_ciabr(env, value1);
813 
814     return H_SUCCESS;
815 }
816 
817 static target_ulong h_set_mode_resource_set_dawr0(PowerPCCPU *cpu,
818                                                   SpaprMachineState *spapr,
819                                                   target_ulong mflags,
820                                                   target_ulong value1,
821                                                   target_ulong value2)
822 {
823     CPUPPCState *env = &cpu->env;
824 
825     assert(tcg_enabled()); /* KVM will have handled this */
826 
827     if (mflags) {
828         return H_UNSUPPORTED_FLAG;
829     }
830     if (value2 & PPC_BIT(61)) {
831         return H_P4;
832     }
833 
834     ppc_store_dawr0(env, value1);
835     ppc_store_dawrx0(env, value2);
836 
837     return H_SUCCESS;
838 }
839 
840 static target_ulong h_set_mode_resource_le(PowerPCCPU *cpu,
841                                            SpaprMachineState *spapr,
842                                            target_ulong mflags,
843                                            target_ulong value1,
844                                            target_ulong value2)
845 {
846     if (value1) {
847         return H_P3;
848     }
849     if (value2) {
850         return H_P4;
851     }
852 
853     switch (mflags) {
854     case H_SET_MODE_ENDIAN_BIG:
855         spapr_set_all_lpcrs(0, LPCR_ILE);
856         spapr_pci_switch_vga(spapr, true);
857         return H_SUCCESS;
858 
859     case H_SET_MODE_ENDIAN_LITTLE:
860         spapr_set_all_lpcrs(LPCR_ILE, LPCR_ILE);
861         spapr_pci_switch_vga(spapr, false);
862         return H_SUCCESS;
863     }
864 
865     return H_UNSUPPORTED_FLAG;
866 }
867 
868 static target_ulong h_set_mode_resource_addr_trans_mode(PowerPCCPU *cpu,
869                                                         SpaprMachineState *spapr,
870                                                         target_ulong mflags,
871                                                         target_ulong value1,
872                                                         target_ulong value2)
873 {
874     if (value1) {
875         return H_P3;
876     }
877 
878     if (value2) {
879         return H_P4;
880     }
881 
882     /*
883      * AIL-1 is not architected, and AIL-2 is not supported by QEMU spapr.
884      * It is supported for faithful emulation of bare metal systems, but for
885      * compatibility concerns we leave it out of the pseries machine.
886      */
887     if (mflags != 0 && mflags != 3) {
888         return H_UNSUPPORTED_FLAG;
889     }
890 
891     if (mflags == 3) {
892         if (!spapr_get_cap(spapr, SPAPR_CAP_AIL_MODE_3)) {
893             return H_UNSUPPORTED_FLAG;
894         }
895     }
896 
897     spapr_set_all_lpcrs(mflags << LPCR_AIL_SHIFT, LPCR_AIL);
898 
899     return H_SUCCESS;
900 }
901 
902 static target_ulong h_set_mode(PowerPCCPU *cpu, SpaprMachineState *spapr,
903                                target_ulong opcode, target_ulong *args)
904 {
905     target_ulong resource = args[1];
906     target_ulong ret = H_P2;
907 
908     switch (resource) {
909     case H_SET_MODE_RESOURCE_SET_CIABR:
910         ret = h_set_mode_resource_set_ciabr(cpu, spapr, args[0], args[2],
911                                             args[3]);
912         break;
913     case H_SET_MODE_RESOURCE_SET_DAWR0:
914         ret = h_set_mode_resource_set_dawr0(cpu, spapr, args[0], args[2],
915                                             args[3]);
916         break;
917     case H_SET_MODE_RESOURCE_LE:
918         ret = h_set_mode_resource_le(cpu, spapr, args[0], args[2], args[3]);
919         break;
920     case H_SET_MODE_RESOURCE_ADDR_TRANS_MODE:
921         ret = h_set_mode_resource_addr_trans_mode(cpu, spapr, args[0],
922                                                   args[2], args[3]);
923         break;
924     }
925 
926     return ret;
927 }
928 
929 static target_ulong h_clean_slb(PowerPCCPU *cpu, SpaprMachineState *spapr,
930                                 target_ulong opcode, target_ulong *args)
931 {
932     qemu_log_mask(LOG_UNIMP, "Unimplemented SPAPR hcall 0x"TARGET_FMT_lx"%s\n",
933                   opcode, " (H_CLEAN_SLB)");
934     return H_FUNCTION;
935 }
936 
937 static target_ulong h_invalidate_pid(PowerPCCPU *cpu, SpaprMachineState *spapr,
938                                      target_ulong opcode, target_ulong *args)
939 {
940     qemu_log_mask(LOG_UNIMP, "Unimplemented SPAPR hcall 0x"TARGET_FMT_lx"%s\n",
941                   opcode, " (H_INVALIDATE_PID)");
942     return H_FUNCTION;
943 }
944 
945 static void spapr_check_setup_free_hpt(SpaprMachineState *spapr,
946                                        uint64_t patbe_old, uint64_t patbe_new)
947 {
948     /*
949      * We have 4 Options:
950      * HASH->HASH || RADIX->RADIX || NOTHING->RADIX : Do Nothing
951      * HASH->RADIX                                  : Free HPT
952      * RADIX->HASH                                  : Allocate HPT
953      * NOTHING->HASH                                : Allocate HPT
954      * Note: NOTHING implies the case where we said the guest could choose
955      *       later and so assumed radix and now it's called H_REG_PROC_TBL
956      */
957 
958     if ((patbe_old & PATE1_GR) == (patbe_new & PATE1_GR)) {
959         /* We assume RADIX, so this catches all the "Do Nothing" cases */
960     } else if (!(patbe_old & PATE1_GR)) {
961         /* HASH->RADIX : Free HPT */
962         spapr_free_hpt(spapr);
963     } else if (!(patbe_new & PATE1_GR)) {
964         /* RADIX->HASH || NOTHING->HASH : Allocate HPT */
965         spapr_setup_hpt(spapr);
966     }
967     return;
968 }
969 
970 #define FLAGS_MASK              0x01FULL
971 #define FLAG_MODIFY             0x10
972 #define FLAG_REGISTER           0x08
973 #define FLAG_RADIX              0x04
974 #define FLAG_HASH_PROC_TBL      0x02
975 #define FLAG_GTSE               0x01
976 
977 static target_ulong h_register_process_table(PowerPCCPU *cpu,
978                                              SpaprMachineState *spapr,
979                                              target_ulong opcode,
980                                              target_ulong *args)
981 {
982     target_ulong flags = args[0];
983     target_ulong proc_tbl = args[1];
984     target_ulong page_size = args[2];
985     target_ulong table_size = args[3];
986     target_ulong update_lpcr = 0;
987     target_ulong table_byte_size;
988     uint64_t cproc;
989 
990     if (flags & ~FLAGS_MASK) { /* Check no reserved bits are set */
991         return H_PARAMETER;
992     }
993     if (flags & FLAG_MODIFY) {
994         if (flags & FLAG_REGISTER) {
995             /* Check process table alignment */
996             table_byte_size = 1ULL << (table_size + 12);
997             if (proc_tbl & (table_byte_size - 1)) {
998                 qemu_log_mask(LOG_GUEST_ERROR,
999                     "%s: process table not properly aligned: proc_tbl 0x"
1000                     TARGET_FMT_lx" proc_tbl_size 0x"TARGET_FMT_lx"\n",
1001                     __func__, proc_tbl, table_byte_size);
1002             }
1003             if (flags & FLAG_RADIX) { /* Register new RADIX process table */
1004                 if (proc_tbl & 0xfff || proc_tbl >> 60) {
1005                     return H_P2;
1006                 } else if (page_size) {
1007                     return H_P3;
1008                 } else if (table_size > 24) {
1009                     return H_P4;
1010                 }
1011                 cproc = PATE1_GR | proc_tbl | table_size;
1012             } else { /* Register new HPT process table */
1013                 if (flags & FLAG_HASH_PROC_TBL) { /* Hash with Segment Tables */
1014                     /* TODO - Not Supported */
1015                     /* Technically caused by flag bits => H_PARAMETER */
1016                     return H_PARAMETER;
1017                 } else { /* Hash with SLB */
1018                     if (proc_tbl >> 38) {
1019                         return H_P2;
1020                     } else if (page_size & ~0x7) {
1021                         return H_P3;
1022                     } else if (table_size > 24) {
1023                         return H_P4;
1024                     }
1025                 }
1026                 cproc = (proc_tbl << 25) | page_size << 5 | table_size;
1027             }
1028 
1029         } else { /* Deregister current process table */
1030             /*
1031              * Set to benign value: (current GR) | 0. This allows
1032              * deregistration in KVM to succeed even if the radix bit
1033              * in flags doesn't match the radix bit in the old PATE.
1034              */
1035             cproc = spapr->patb_entry & PATE1_GR;
1036         }
1037     } else { /* Maintain current registration */
1038         if (!(flags & FLAG_RADIX) != !(spapr->patb_entry & PATE1_GR)) {
1039             /* Technically caused by flag bits => H_PARAMETER */
1040             return H_PARAMETER; /* Existing Process Table Mismatch */
1041         }
1042         cproc = spapr->patb_entry;
1043     }
1044 
1045     /* Check if we need to setup OR free the hpt */
1046     spapr_check_setup_free_hpt(spapr, spapr->patb_entry, cproc);
1047 
1048     spapr->patb_entry = cproc; /* Save new process table */
1049 
1050     /* Update the UPRT, HR and GTSE bits in the LPCR for all cpus */
1051     if (flags & FLAG_RADIX)     /* Radix must use process tables, also set HR */
1052         update_lpcr |= (LPCR_UPRT | LPCR_HR);
1053     else if (flags & FLAG_HASH_PROC_TBL) /* Hash with process tables */
1054         update_lpcr |= LPCR_UPRT;
1055     if (flags & FLAG_GTSE)      /* Guest translation shootdown enable */
1056         update_lpcr |= LPCR_GTSE;
1057 
1058     spapr_set_all_lpcrs(update_lpcr, LPCR_UPRT | LPCR_HR | LPCR_GTSE);
1059 
1060     if (kvm_enabled()) {
1061         return kvmppc_configure_v3_mmu(cpu, flags & FLAG_RADIX,
1062                                        flags & FLAG_GTSE, cproc);
1063     }
1064     return H_SUCCESS;
1065 }
1066 
1067 #define H_SIGNAL_SYS_RESET_ALL         -1
1068 #define H_SIGNAL_SYS_RESET_ALLBUTSELF  -2
1069 
1070 static target_ulong h_signal_sys_reset(PowerPCCPU *cpu,
1071                                        SpaprMachineState *spapr,
1072                                        target_ulong opcode, target_ulong *args)
1073 {
1074     target_long target = args[0];
1075     CPUState *cs;
1076 
1077     if (target < 0) {
1078         /* Broadcast */
1079         if (target < H_SIGNAL_SYS_RESET_ALLBUTSELF) {
1080             return H_PARAMETER;
1081         }
1082 
1083         CPU_FOREACH(cs) {
1084             PowerPCCPU *c = POWERPC_CPU(cs);
1085 
1086             if (target == H_SIGNAL_SYS_RESET_ALLBUTSELF) {
1087                 if (c == cpu) {
1088                     continue;
1089                 }
1090             }
1091             run_on_cpu(cs, spapr_do_system_reset_on_cpu, RUN_ON_CPU_NULL);
1092         }
1093         return H_SUCCESS;
1094 
1095     } else {
1096         /* Unicast */
1097         cs = CPU(spapr_find_cpu(target));
1098         if (cs) {
1099             run_on_cpu(cs, spapr_do_system_reset_on_cpu, RUN_ON_CPU_NULL);
1100             return H_SUCCESS;
1101         }
1102         return H_PARAMETER;
1103     }
1104 }
1105 
1106 /* Returns either a logical PVR or zero if none was found */
1107 static uint32_t cas_check_pvr(PowerPCCPU *cpu, uint32_t max_compat,
1108                               target_ulong *addr, bool *raw_mode_supported)
1109 {
1110     bool explicit_match = false; /* Matched the CPU's real PVR */
1111     uint32_t best_compat = 0;
1112     int i;
1113 
1114     /*
1115      * We scan the supplied table of PVRs looking for two things
1116      *   1. Is our real CPU PVR in the list?
1117      *   2. What's the "best" listed logical PVR
1118      */
1119     for (i = 0; i < 512; ++i) {
1120         uint32_t pvr, pvr_mask;
1121 
1122         pvr_mask = ldl_be_phys(&address_space_memory, *addr);
1123         pvr = ldl_be_phys(&address_space_memory, *addr + 4);
1124         *addr += 8;
1125 
1126         if (~pvr_mask & pvr) {
1127             break; /* Terminator record */
1128         }
1129 
1130         if ((cpu->env.spr[SPR_PVR] & pvr_mask) == (pvr & pvr_mask)) {
1131             explicit_match = true;
1132         } else {
1133             if (ppc_check_compat(cpu, pvr, best_compat, max_compat)) {
1134                 best_compat = pvr;
1135             }
1136         }
1137     }
1138 
1139     *raw_mode_supported = explicit_match;
1140 
1141     /* Parsing finished */
1142     trace_spapr_cas_pvr(cpu->compat_pvr, explicit_match, best_compat);
1143 
1144     return best_compat;
1145 }
1146 
1147 static
1148 target_ulong do_client_architecture_support(PowerPCCPU *cpu,
1149                                             SpaprMachineState *spapr,
1150                                             target_ulong vec,
1151                                             target_ulong fdt_bufsize)
1152 {
1153     target_ulong ov_table; /* Working address in data buffer */
1154     uint32_t cas_pvr;
1155     SpaprOptionVector *ov1_guest, *ov5_guest;
1156     bool guest_radix;
1157     bool raw_mode_supported = false;
1158     bool guest_xive;
1159     CPUState *cs;
1160     void *fdt;
1161     uint32_t max_compat = spapr->max_compat_pvr;
1162 
1163     /* CAS is supposed to be called early when only the boot vCPU is active. */
1164     CPU_FOREACH(cs) {
1165         if (cs == CPU(cpu)) {
1166             continue;
1167         }
1168         if (!cs->halted) {
1169             warn_report("guest has multiple active vCPUs at CAS, which is not allowed");
1170             return H_MULTI_THREADS_ACTIVE;
1171         }
1172     }
1173 
1174     cas_pvr = cas_check_pvr(cpu, max_compat, &vec, &raw_mode_supported);
1175     if (!cas_pvr && (!raw_mode_supported || max_compat)) {
1176         /*
1177          * We couldn't find a suitable compatibility mode, and either
1178          * the guest doesn't support "raw" mode for this CPU, or "raw"
1179          * mode is disabled because a maximum compat mode is set.
1180          */
1181         error_report("Couldn't negotiate a suitable PVR during CAS");
1182         return H_HARDWARE;
1183     }
1184 
1185     /* Update CPUs */
1186     if (cpu->compat_pvr != cas_pvr) {
1187         Error *local_err = NULL;
1188 
1189         if (ppc_set_compat_all(cas_pvr, &local_err) < 0) {
1190             /* We fail to set compat mode (likely because running with KVM PR),
1191              * but maybe we can fallback to raw mode if the guest supports it.
1192              */
1193             if (!raw_mode_supported) {
1194                 error_report_err(local_err);
1195                 return H_HARDWARE;
1196             }
1197             error_free(local_err);
1198         }
1199     }
1200 
1201     /* For the future use: here @ov_table points to the first option vector */
1202     ov_table = vec;
1203 
1204     ov1_guest = spapr_ovec_parse_vector(ov_table, 1);
1205     if (!ov1_guest) {
1206         warn_report("guest didn't provide option vector 1");
1207         return H_PARAMETER;
1208     }
1209     ov5_guest = spapr_ovec_parse_vector(ov_table, 5);
1210     if (!ov5_guest) {
1211         spapr_ovec_cleanup(ov1_guest);
1212         warn_report("guest didn't provide option vector 5");
1213         return H_PARAMETER;
1214     }
1215     if (spapr_ovec_test(ov5_guest, OV5_MMU_BOTH)) {
1216         error_report("guest requested hash and radix MMU, which is invalid.");
1217         exit(EXIT_FAILURE);
1218     }
1219     if (spapr_ovec_test(ov5_guest, OV5_XIVE_BOTH)) {
1220         error_report("guest requested an invalid interrupt mode");
1221         exit(EXIT_FAILURE);
1222     }
1223 
1224     guest_radix = spapr_ovec_test(ov5_guest, OV5_MMU_RADIX_300);
1225 
1226     guest_xive = spapr_ovec_test(ov5_guest, OV5_XIVE_EXPLOIT);
1227 
1228     /*
1229      * HPT resizing is a bit of a special case, because when enabled
1230      * we assume an HPT guest will support it until it says it
1231      * doesn't, instead of assuming it won't support it until it says
1232      * it does.  Strictly speaking that approach could break for
1233      * guests which don't make a CAS call, but those are so old we
1234      * don't care about them.  Without that assumption we'd have to
1235      * make at least a temporary allocation of an HPT sized for max
1236      * memory, which could be impossibly difficult under KVM HV if
1237      * maxram is large.
1238      */
1239     if (!guest_radix && !spapr_ovec_test(ov5_guest, OV5_HPT_RESIZE)) {
1240         int maxshift = spapr_hpt_shift_for_ramsize(MACHINE(spapr)->maxram_size);
1241 
1242         if (spapr->resize_hpt == SPAPR_RESIZE_HPT_REQUIRED) {
1243             error_report(
1244                 "h_client_architecture_support: Guest doesn't support HPT resizing, but resize-hpt=required");
1245             exit(1);
1246         }
1247 
1248         if (spapr->htab_shift < maxshift) {
1249             /* Guest doesn't know about HPT resizing, so we
1250              * pre-emptively resize for the maximum permitted RAM.  At
1251              * the point this is called, nothing should have been
1252              * entered into the existing HPT */
1253             spapr_reallocate_hpt(spapr, maxshift, &error_fatal);
1254             push_sregs_to_kvm_pr(spapr);
1255         }
1256     }
1257 
1258     /* NOTE: there are actually a number of ov5 bits where input from the
1259      * guest is always zero, and the platform/QEMU enables them independently
1260      * of guest input. To model these properly we'd want some sort of mask,
1261      * but since they only currently apply to memory migration as defined
1262      * by LoPAPR 1.1, 14.5.4.8, which QEMU doesn't implement, we don't need
1263      * to worry about this for now.
1264      */
1265 
1266     /* full range of negotiated ov5 capabilities */
1267     spapr_ovec_intersect(spapr->ov5_cas, spapr->ov5, ov5_guest);
1268     spapr_ovec_cleanup(ov5_guest);
1269 
1270     spapr_check_mmu_mode(guest_radix);
1271 
1272     spapr->cas_pre_isa3_guest = !spapr_ovec_test(ov1_guest, OV1_PPC_3_00);
1273     spapr_ovec_cleanup(ov1_guest);
1274 
1275     /*
1276      * Check for NUMA affinity conditions now that we know which NUMA
1277      * affinity the guest will use.
1278      */
1279     spapr_numa_associativity_check(spapr);
1280 
1281     /*
1282      * Ensure the guest asks for an interrupt mode we support;
1283      * otherwise terminate the boot.
1284      */
1285     if (guest_xive) {
1286         if (!spapr->irq->xive) {
1287             error_report(
1288 "Guest requested unavailable interrupt mode (XIVE), try the ic-mode=xive or ic-mode=dual machine property");
1289             exit(EXIT_FAILURE);
1290         }
1291     } else {
1292         if (!spapr->irq->xics) {
1293             error_report(
1294 "Guest requested unavailable interrupt mode (XICS), either don't set the ic-mode machine property or try ic-mode=xics or ic-mode=dual");
1295             exit(EXIT_FAILURE);
1296         }
1297     }
1298 
1299     spapr_irq_update_active_intc(spapr);
1300 
1301     /*
1302      * Process all pending hot-plug/unplug requests now. An updated full
1303      * rendered FDT will be returned to the guest.
1304      */
1305     spapr_drc_reset_all(spapr);
1306     spapr_clear_pending_hotplug_events(spapr);
1307 
1308     /*
1309      * If spapr_machine_reset() did not set up a HPT but one is necessary
1310      * (because the guest isn't going to use radix) then set it up here.
1311      */
1312     if ((spapr->patb_entry & PATE1_GR) && !guest_radix) {
1313         /* legacy hash or new hash: */
1314         spapr_setup_hpt(spapr);
1315     }
1316 
1317     fdt = spapr_build_fdt(spapr, spapr->vof != NULL, fdt_bufsize);
1318     g_free(spapr->fdt_blob);
1319     spapr->fdt_size = fdt_totalsize(fdt);
1320     spapr->fdt_initial_size = spapr->fdt_size;
1321     spapr->fdt_blob = fdt;
1322 
1323     /*
1324      * Set the machine->fdt pointer again since we just freed
1325      * it above (by freeing spapr->fdt_blob). We set this
1326      * pointer to enable support for the 'dumpdtb' QMP/HMP
1327      * command.
1328      */
1329     MACHINE(spapr)->fdt = fdt;
1330 
1331     return H_SUCCESS;
1332 }
1333 
1334 static target_ulong h_client_architecture_support(PowerPCCPU *cpu,
1335                                                   SpaprMachineState *spapr,
1336                                                   target_ulong opcode,
1337                                                   target_ulong *args)
1338 {
1339     target_ulong vec = ppc64_phys_to_real(args[0]);
1340     target_ulong fdt_buf = args[1];
1341     target_ulong fdt_bufsize = args[2];
1342     target_ulong ret;
1343     SpaprDeviceTreeUpdateHeader hdr = { .version_id = 1 };
1344 
1345     if (fdt_bufsize < sizeof(hdr)) {
1346         error_report("SLOF provided insufficient CAS buffer "
1347                      TARGET_FMT_lu " (min: %zu)", fdt_bufsize, sizeof(hdr));
1348         exit(EXIT_FAILURE);
1349     }
1350 
1351     fdt_bufsize -= sizeof(hdr);
1352 
1353     ret = do_client_architecture_support(cpu, spapr, vec, fdt_bufsize);
1354     if (ret == H_SUCCESS) {
1355         _FDT((fdt_pack(spapr->fdt_blob)));
1356         spapr->fdt_size = fdt_totalsize(spapr->fdt_blob);
1357         spapr->fdt_initial_size = spapr->fdt_size;
1358 
1359         cpu_physical_memory_write(fdt_buf, &hdr, sizeof(hdr));
1360         cpu_physical_memory_write(fdt_buf + sizeof(hdr), spapr->fdt_blob,
1361                                   spapr->fdt_size);
1362         trace_spapr_cas_continue(spapr->fdt_size + sizeof(hdr));
1363     }
1364 
1365     return ret;
1366 }
1367 
1368 target_ulong spapr_vof_client_architecture_support(MachineState *ms,
1369                                                    CPUState *cs,
1370                                                    target_ulong ovec_addr)
1371 {
1372     SpaprMachineState *spapr = SPAPR_MACHINE(ms);
1373 
1374     target_ulong ret = do_client_architecture_support(POWERPC_CPU(cs), spapr,
1375                                                       ovec_addr, FDT_MAX_SIZE);
1376 
1377     /*
1378      * This adds stdout and generates phandles for boottime and CAS FDTs.
1379      * It is alright to update the FDT here as do_client_architecture_support()
1380      * does not pack it.
1381      */
1382     spapr_vof_client_dt_finalize(spapr, spapr->fdt_blob);
1383 
1384     return ret;
1385 }
1386 
1387 static target_ulong h_get_cpu_characteristics(PowerPCCPU *cpu,
1388                                               SpaprMachineState *spapr,
1389                                               target_ulong opcode,
1390                                               target_ulong *args)
1391 {
1392     uint64_t characteristics = H_CPU_CHAR_HON_BRANCH_HINTS &
1393                                ~H_CPU_CHAR_THR_RECONF_TRIG;
1394     uint64_t behaviour = H_CPU_BEHAV_FAVOUR_SECURITY;
1395     uint8_t safe_cache = spapr_get_cap(spapr, SPAPR_CAP_CFPC);
1396     uint8_t safe_bounds_check = spapr_get_cap(spapr, SPAPR_CAP_SBBC);
1397     uint8_t safe_indirect_branch = spapr_get_cap(spapr, SPAPR_CAP_IBS);
1398     uint8_t count_cache_flush_assist = spapr_get_cap(spapr,
1399                                                      SPAPR_CAP_CCF_ASSIST);
1400 
1401     switch (safe_cache) {
1402     case SPAPR_CAP_WORKAROUND:
1403         characteristics |= H_CPU_CHAR_L1D_FLUSH_ORI30;
1404         characteristics |= H_CPU_CHAR_L1D_FLUSH_TRIG2;
1405         characteristics |= H_CPU_CHAR_L1D_THREAD_PRIV;
1406         behaviour |= H_CPU_BEHAV_L1D_FLUSH_PR;
1407         break;
1408     case SPAPR_CAP_FIXED:
1409         behaviour |= H_CPU_BEHAV_NO_L1D_FLUSH_ENTRY;
1410         behaviour |= H_CPU_BEHAV_NO_L1D_FLUSH_UACCESS;
1411         break;
1412     default: /* broken */
1413         assert(safe_cache == SPAPR_CAP_BROKEN);
1414         behaviour |= H_CPU_BEHAV_L1D_FLUSH_PR;
1415         break;
1416     }
1417 
1418     switch (safe_bounds_check) {
1419     case SPAPR_CAP_WORKAROUND:
1420         characteristics |= H_CPU_CHAR_SPEC_BAR_ORI31;
1421         behaviour |= H_CPU_BEHAV_BNDS_CHK_SPEC_BAR;
1422         break;
1423     case SPAPR_CAP_FIXED:
1424         break;
1425     default: /* broken */
1426         assert(safe_bounds_check == SPAPR_CAP_BROKEN);
1427         behaviour |= H_CPU_BEHAV_BNDS_CHK_SPEC_BAR;
1428         break;
1429     }
1430 
1431     switch (safe_indirect_branch) {
1432     case SPAPR_CAP_FIXED_NA:
1433         break;
1434     case SPAPR_CAP_FIXED_CCD:
1435         characteristics |= H_CPU_CHAR_CACHE_COUNT_DIS;
1436         break;
1437     case SPAPR_CAP_FIXED_IBS:
1438         characteristics |= H_CPU_CHAR_BCCTRL_SERIALISED;
1439         break;
1440     case SPAPR_CAP_WORKAROUND:
1441         behaviour |= H_CPU_BEHAV_FLUSH_COUNT_CACHE;
1442         if (count_cache_flush_assist) {
1443             characteristics |= H_CPU_CHAR_BCCTR_FLUSH_ASSIST;
1444         }
1445         break;
1446     default: /* broken */
1447         assert(safe_indirect_branch == SPAPR_CAP_BROKEN);
1448         break;
1449     }
1450 
1451     args[0] = characteristics;
1452     args[1] = behaviour;
1453     return H_SUCCESS;
1454 }
1455 
1456 static target_ulong h_update_dt(PowerPCCPU *cpu, SpaprMachineState *spapr,
1457                                 target_ulong opcode, target_ulong *args)
1458 {
1459     target_ulong dt = ppc64_phys_to_real(args[0]);
1460     struct fdt_header hdr = { 0 };
1461     unsigned cb;
1462     SpaprMachineClass *smc = SPAPR_MACHINE_GET_CLASS(spapr);
1463     void *fdt;
1464 
1465     cpu_physical_memory_read(dt, &hdr, sizeof(hdr));
1466     cb = fdt32_to_cpu(hdr.totalsize);
1467 
1468     if (!smc->update_dt_enabled) {
1469         return H_SUCCESS;
1470     }
1471 
1472     /* Check that the fdt did not grow out of proportion */
1473     if (cb > spapr->fdt_initial_size * 2) {
1474         trace_spapr_update_dt_failed_size(spapr->fdt_initial_size, cb,
1475                                           fdt32_to_cpu(hdr.magic));
1476         return H_PARAMETER;
1477     }
1478 
1479     fdt = g_malloc0(cb);
1480     cpu_physical_memory_read(dt, fdt, cb);
1481 
1482     /* Check the fdt consistency */
1483     if (fdt_check_full(fdt, cb)) {
1484         trace_spapr_update_dt_failed_check(spapr->fdt_initial_size, cb,
1485                                            fdt32_to_cpu(hdr.magic));
1486         return H_PARAMETER;
1487     }
1488 
1489     g_free(spapr->fdt_blob);
1490     spapr->fdt_size = cb;
1491     spapr->fdt_blob = fdt;
1492     trace_spapr_update_dt(cb);
1493 
1494     return H_SUCCESS;
1495 }
1496 
1497 static spapr_hcall_fn papr_hypercall_table[(MAX_HCALL_OPCODE / 4) + 1];
1498 static spapr_hcall_fn kvmppc_hypercall_table[KVMPPC_HCALL_MAX - KVMPPC_HCALL_BASE + 1];
1499 static spapr_hcall_fn svm_hypercall_table[(SVM_HCALL_MAX - SVM_HCALL_BASE) / 4 + 1];
1500 
1501 void spapr_register_hypercall(target_ulong opcode, spapr_hcall_fn fn)
1502 {
1503     spapr_hcall_fn *slot;
1504 
1505     if (opcode <= MAX_HCALL_OPCODE) {
1506         assert((opcode & 0x3) == 0);
1507 
1508         slot = &papr_hypercall_table[opcode / 4];
1509     } else if (opcode >= SVM_HCALL_BASE && opcode <= SVM_HCALL_MAX) {
1510         /* we only have SVM-related hcall numbers assigned in multiples of 4 */
1511         assert((opcode & 0x3) == 0);
1512 
1513         slot = &svm_hypercall_table[(opcode - SVM_HCALL_BASE) / 4];
1514     } else {
1515         assert((opcode >= KVMPPC_HCALL_BASE) && (opcode <= KVMPPC_HCALL_MAX));
1516 
1517         slot = &kvmppc_hypercall_table[opcode - KVMPPC_HCALL_BASE];
1518     }
1519 
1520     assert(!(*slot));
1521     *slot = fn;
1522 }
1523 
1524 target_ulong spapr_hypercall(PowerPCCPU *cpu, target_ulong opcode,
1525                              target_ulong *args)
1526 {
1527     SpaprMachineState *spapr = SPAPR_MACHINE(qdev_get_machine());
1528 
1529     if ((opcode <= MAX_HCALL_OPCODE)
1530         && ((opcode & 0x3) == 0)) {
1531         spapr_hcall_fn fn = papr_hypercall_table[opcode / 4];
1532 
1533         if (fn) {
1534             return fn(cpu, spapr, opcode, args);
1535         }
1536     } else if ((opcode >= SVM_HCALL_BASE) &&
1537                (opcode <= SVM_HCALL_MAX)) {
1538         spapr_hcall_fn fn = svm_hypercall_table[(opcode - SVM_HCALL_BASE) / 4];
1539 
1540         if (fn) {
1541             return fn(cpu, spapr, opcode, args);
1542         }
1543     } else if ((opcode >= KVMPPC_HCALL_BASE) &&
1544                (opcode <= KVMPPC_HCALL_MAX)) {
1545         spapr_hcall_fn fn = kvmppc_hypercall_table[opcode - KVMPPC_HCALL_BASE];
1546 
1547         if (fn) {
1548             return fn(cpu, spapr, opcode, args);
1549         }
1550     }
1551 
1552     qemu_log_mask(LOG_UNIMP, "Unimplemented SPAPR hcall 0x" TARGET_FMT_lx "\n",
1553                   opcode);
1554     return H_FUNCTION;
1555 }
1556 
1557 #ifdef CONFIG_TCG
1558 static void hypercall_register_softmmu(void)
1559 {
1560     /* DO NOTHING */
1561 }
1562 #else
1563 static target_ulong h_softmmu(PowerPCCPU *cpu, SpaprMachineState *spapr,
1564                             target_ulong opcode, target_ulong *args)
1565 {
1566     g_assert_not_reached();
1567 }
1568 
1569 static void hypercall_register_softmmu(void)
1570 {
1571     /* hcall-pft */
1572     spapr_register_hypercall(H_ENTER, h_softmmu);
1573     spapr_register_hypercall(H_REMOVE, h_softmmu);
1574     spapr_register_hypercall(H_PROTECT, h_softmmu);
1575     spapr_register_hypercall(H_READ, h_softmmu);
1576 
1577     /* hcall-bulk */
1578     spapr_register_hypercall(H_BULK_REMOVE, h_softmmu);
1579 }
1580 #endif
1581 
1582 static void hypercall_register_types(void)
1583 {
1584     hypercall_register_softmmu();
1585 
1586     /* hcall-hpt-resize */
1587     spapr_register_hypercall(H_RESIZE_HPT_PREPARE, h_resize_hpt_prepare);
1588     spapr_register_hypercall(H_RESIZE_HPT_COMMIT, h_resize_hpt_commit);
1589 
1590     /* hcall-splpar */
1591     spapr_register_hypercall(H_REGISTER_VPA, h_register_vpa);
1592     spapr_register_hypercall(H_CEDE, h_cede);
1593     spapr_register_hypercall(H_CONFER, h_confer);
1594     spapr_register_hypercall(H_PROD, h_prod);
1595 
1596     /* hcall-join */
1597     spapr_register_hypercall(H_JOIN, h_join);
1598 
1599     spapr_register_hypercall(H_SIGNAL_SYS_RESET, h_signal_sys_reset);
1600 
1601     /* processor register resource access h-calls */
1602     spapr_register_hypercall(H_SET_SPRG0, h_set_sprg0);
1603     spapr_register_hypercall(H_SET_DABR, h_set_dabr);
1604     spapr_register_hypercall(H_SET_XDABR, h_set_xdabr);
1605     spapr_register_hypercall(H_PAGE_INIT, h_page_init);
1606     spapr_register_hypercall(H_SET_MODE, h_set_mode);
1607 
1608     /* In Memory Table MMU h-calls */
1609     spapr_register_hypercall(H_CLEAN_SLB, h_clean_slb);
1610     spapr_register_hypercall(H_INVALIDATE_PID, h_invalidate_pid);
1611     spapr_register_hypercall(H_REGISTER_PROC_TBL, h_register_process_table);
1612 
1613     /* hcall-get-cpu-characteristics */
1614     spapr_register_hypercall(H_GET_CPU_CHARACTERISTICS,
1615                              h_get_cpu_characteristics);
1616 
1617     /* "debugger" hcalls (also used by SLOF). Note: We do -not- differentiate
1618      * here between the "CI" and the "CACHE" variants, they will use whatever
1619      * mapping attributes qemu is using. When using KVM, the kernel will
1620      * enforce the attributes more strongly
1621      */
1622     spapr_register_hypercall(H_LOGICAL_CI_LOAD, h_logical_load);
1623     spapr_register_hypercall(H_LOGICAL_CI_STORE, h_logical_store);
1624     spapr_register_hypercall(H_LOGICAL_CACHE_LOAD, h_logical_load);
1625     spapr_register_hypercall(H_LOGICAL_CACHE_STORE, h_logical_store);
1626     spapr_register_hypercall(H_LOGICAL_ICBI, h_logical_icbi);
1627     spapr_register_hypercall(H_LOGICAL_DCBF, h_logical_dcbf);
1628     spapr_register_hypercall(KVMPPC_H_LOGICAL_MEMOP, h_logical_memop);
1629 
1630     /* qemu/KVM-PPC specific hcalls */
1631     spapr_register_hypercall(KVMPPC_H_RTAS, h_rtas);
1632 
1633     /* ibm,client-architecture-support support */
1634     spapr_register_hypercall(KVMPPC_H_CAS, h_client_architecture_support);
1635 
1636     spapr_register_hypercall(KVMPPC_H_UPDATE_DT, h_update_dt);
1637 
1638     spapr_register_nested();
1639 }
1640 
1641 type_init(hypercall_register_types)
1642