xref: /qemu/hw/ppc/spapr_hcall.c (revision 3e84d0381578467931d1145a32a55b7e1c3b2b9e)
1 #include "qemu/osdep.h"
2 #include "qemu/cutils.h"
3 #include "qapi/error.h"
4 #include "system/hw_accel.h"
5 #include "system/runstate.h"
6 #include "system/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     } else if (tcg_enabled()) {
127         return vhyp_mmu_resize_hpt_prepare(cpu, spapr, shift);
128     } else {
129         g_assert_not_reached();
130     }
131 }
132 
133 static void do_push_sregs_to_kvm_pr(CPUState *cs, run_on_cpu_data data)
134 {
135     int ret;
136 
137     cpu_synchronize_state(cs);
138 
139     ret = kvmppc_put_books_sregs(POWERPC_CPU(cs));
140     if (ret < 0) {
141         error_report("failed to push sregs to KVM: %s", strerror(-ret));
142         exit(1);
143     }
144 }
145 
146 void push_sregs_to_kvm_pr(SpaprMachineState *spapr)
147 {
148     CPUState *cs;
149 
150     /*
151      * This is a hack for the benefit of KVM PR - it abuses the SDR1
152      * slot in kvm_sregs to communicate the userspace address of the
153      * HPT
154      */
155     if (!kvm_enabled() || !spapr->htab) {
156         return;
157     }
158 
159     CPU_FOREACH(cs) {
160         run_on_cpu(cs, do_push_sregs_to_kvm_pr, RUN_ON_CPU_NULL);
161     }
162 }
163 
164 static target_ulong h_resize_hpt_commit(PowerPCCPU *cpu,
165                                         SpaprMachineState *spapr,
166                                         target_ulong opcode,
167                                         target_ulong *args)
168 {
169     target_ulong flags = args[0];
170     target_ulong shift = args[1];
171     int rc;
172 
173     if (spapr->resize_hpt == SPAPR_RESIZE_HPT_DISABLED) {
174         return H_AUTHORITY;
175     }
176 
177     if (!spapr->htab_shift) {
178         /* Radix guest, no HPT */
179         return H_NOT_AVAILABLE;
180     }
181 
182     trace_spapr_h_resize_hpt_commit(flags, shift);
183 
184     rc = kvmppc_resize_hpt_commit(cpu, flags, shift);
185     if (rc != -ENOSYS) {
186         rc = resize_hpt_convert_rc(rc);
187         if (rc == H_SUCCESS) {
188             /* Need to set the new htab_shift in the machine state */
189             spapr->htab_shift = shift;
190         }
191         return rc;
192     }
193 
194     if (kvm_enabled()) {
195         return H_HARDWARE;
196     } else if (tcg_enabled()) {
197         return vhyp_mmu_resize_hpt_commit(cpu, spapr, flags, shift);
198     } else {
199         g_assert_not_reached();
200     }
201 }
202 
203 
204 
205 static target_ulong h_set_sprg0(PowerPCCPU *cpu, SpaprMachineState *spapr,
206                                 target_ulong opcode, target_ulong *args)
207 {
208     cpu_synchronize_state(CPU(cpu));
209     cpu->env.spr[SPR_SPRG0] = args[0];
210 
211     return H_SUCCESS;
212 }
213 
214 static target_ulong h_set_dabr(PowerPCCPU *cpu, SpaprMachineState *spapr,
215                                target_ulong opcode, target_ulong *args)
216 {
217     if (!ppc_has_spr(cpu, SPR_DABR)) {
218         return H_HARDWARE;              /* DABR register not available */
219     }
220     cpu_synchronize_state(CPU(cpu));
221 
222     if (ppc_has_spr(cpu, SPR_DABRX)) {
223         cpu->env.spr[SPR_DABRX] = 0x3;  /* Use Problem and Privileged state */
224     } else if (!(args[0] & 0x4)) {      /* Breakpoint Translation set? */
225         return H_RESERVED_DABR;
226     }
227 
228     cpu->env.spr[SPR_DABR] = args[0];
229     return H_SUCCESS;
230 }
231 
232 static target_ulong h_set_xdabr(PowerPCCPU *cpu, SpaprMachineState *spapr,
233                                 target_ulong opcode, target_ulong *args)
234 {
235     target_ulong dabrx = args[1];
236 
237     if (!ppc_has_spr(cpu, SPR_DABR) || !ppc_has_spr(cpu, SPR_DABRX)) {
238         return H_HARDWARE;
239     }
240 
241     if ((dabrx & ~0xfULL) != 0 || (dabrx & H_DABRX_HYPERVISOR) != 0
242         || (dabrx & (H_DABRX_KERNEL | H_DABRX_USER)) == 0) {
243         return H_PARAMETER;
244     }
245 
246     cpu_synchronize_state(CPU(cpu));
247     cpu->env.spr[SPR_DABRX] = dabrx;
248     cpu->env.spr[SPR_DABR] = args[0];
249 
250     return H_SUCCESS;
251 }
252 
253 static target_ulong h_page_init(PowerPCCPU *cpu, SpaprMachineState *spapr,
254                                 target_ulong opcode, target_ulong *args)
255 {
256     target_ulong flags = args[0];
257     hwaddr dst = args[1];
258     hwaddr src = args[2];
259     hwaddr len = TARGET_PAGE_SIZE;
260     uint8_t *pdst, *psrc;
261     target_long ret = H_SUCCESS;
262 
263     if (flags & ~(H_ICACHE_SYNCHRONIZE | H_ICACHE_INVALIDATE
264                   | H_COPY_PAGE | H_ZERO_PAGE)) {
265         qemu_log_mask(LOG_UNIMP, "h_page_init: Bad flags (" TARGET_FMT_lx "\n",
266                       flags);
267         return H_PARAMETER;
268     }
269 
270     /* Map-in destination */
271     if (!is_ram_address(spapr, dst) || (dst & ~TARGET_PAGE_MASK) != 0) {
272         return H_PARAMETER;
273     }
274     pdst = cpu_physical_memory_map(dst, &len, true);
275     if (!pdst || len != TARGET_PAGE_SIZE) {
276         return H_PARAMETER;
277     }
278 
279     if (flags & H_COPY_PAGE) {
280         /* Map-in source, copy to destination, and unmap source again */
281         if (!is_ram_address(spapr, src) || (src & ~TARGET_PAGE_MASK) != 0) {
282             ret = H_PARAMETER;
283             goto unmap_out;
284         }
285         psrc = cpu_physical_memory_map(src, &len, false);
286         if (!psrc || len != TARGET_PAGE_SIZE) {
287             ret = H_PARAMETER;
288             goto unmap_out;
289         }
290         memcpy(pdst, psrc, len);
291         cpu_physical_memory_unmap(psrc, len, 0, len);
292     } else if (flags & H_ZERO_PAGE) {
293         memset(pdst, 0, len);          /* Just clear the destination page */
294     }
295 
296     if (kvm_enabled() && (flags & H_ICACHE_SYNCHRONIZE) != 0) {
297         kvmppc_dcbst_range(cpu, pdst, len);
298     }
299     if (flags & (H_ICACHE_SYNCHRONIZE | H_ICACHE_INVALIDATE)) {
300         if (kvm_enabled()) {
301             kvmppc_icbi_range(cpu, pdst, len);
302         } else if (tcg_enabled()) {
303             tb_flush(CPU(cpu));
304         } else {
305             g_assert_not_reached();
306         }
307     }
308 
309 unmap_out:
310     cpu_physical_memory_unmap(pdst, TARGET_PAGE_SIZE, 1, len);
311     return ret;
312 }
313 
314 #define FLAGS_REGISTER_VPA         0x0000200000000000ULL
315 #define FLAGS_REGISTER_DTL         0x0000400000000000ULL
316 #define FLAGS_REGISTER_SLBSHADOW   0x0000600000000000ULL
317 #define FLAGS_DEREGISTER_VPA       0x0000a00000000000ULL
318 #define FLAGS_DEREGISTER_DTL       0x0000c00000000000ULL
319 #define FLAGS_DEREGISTER_SLBSHADOW 0x0000e00000000000ULL
320 
321 static target_ulong register_vpa(PowerPCCPU *cpu, target_ulong vpa)
322 {
323     CPUState *cs = CPU(cpu);
324     CPUPPCState *env = &cpu->env;
325     SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
326     uint16_t size;
327     uint8_t tmp;
328 
329     if (vpa == 0) {
330         hcall_dprintf("Can't cope with registering a VPA at logical 0\n");
331         return H_HARDWARE;
332     }
333 
334     if (vpa % env->dcache_line_size) {
335         return H_PARAMETER;
336     }
337     /* FIXME: bounds check the address */
338 
339     size = lduw_be_phys(cs->as, vpa + 0x4);
340 
341     if (size < VPA_MIN_SIZE) {
342         return H_PARAMETER;
343     }
344 
345     /* VPA is not allowed to cross a page boundary */
346     if ((vpa / 4096) != ((vpa + size - 1) / 4096)) {
347         return H_PARAMETER;
348     }
349 
350     spapr_cpu->vpa_addr = vpa;
351 
352     tmp = ldub_phys(cs->as, spapr_cpu->vpa_addr + VPA_SHARED_PROC_OFFSET);
353     tmp |= VPA_SHARED_PROC_VAL;
354     stb_phys(cs->as, spapr_cpu->vpa_addr + VPA_SHARED_PROC_OFFSET, tmp);
355 
356     return H_SUCCESS;
357 }
358 
359 static target_ulong deregister_vpa(PowerPCCPU *cpu, target_ulong vpa)
360 {
361     SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
362 
363     if (spapr_cpu->slb_shadow_addr) {
364         return H_RESOURCE;
365     }
366 
367     if (spapr_cpu->dtl_addr) {
368         return H_RESOURCE;
369     }
370 
371     spapr_cpu->vpa_addr = 0;
372     return H_SUCCESS;
373 }
374 
375 static target_ulong register_slb_shadow(PowerPCCPU *cpu, target_ulong addr)
376 {
377     SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
378     uint32_t size;
379 
380     if (addr == 0) {
381         hcall_dprintf("Can't cope with SLB shadow at logical 0\n");
382         return H_HARDWARE;
383     }
384 
385     size = ldl_be_phys(CPU(cpu)->as, addr + 0x4);
386     if (size < 0x8) {
387         return H_PARAMETER;
388     }
389 
390     if ((addr / 4096) != ((addr + size - 1) / 4096)) {
391         return H_PARAMETER;
392     }
393 
394     if (!spapr_cpu->vpa_addr) {
395         return H_RESOURCE;
396     }
397 
398     spapr_cpu->slb_shadow_addr = addr;
399     spapr_cpu->slb_shadow_size = size;
400 
401     return H_SUCCESS;
402 }
403 
404 static target_ulong deregister_slb_shadow(PowerPCCPU *cpu, target_ulong addr)
405 {
406     SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
407 
408     spapr_cpu->slb_shadow_addr = 0;
409     spapr_cpu->slb_shadow_size = 0;
410     return H_SUCCESS;
411 }
412 
413 static target_ulong register_dtl(PowerPCCPU *cpu, target_ulong addr)
414 {
415     SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
416     uint32_t size;
417 
418     if (addr == 0) {
419         hcall_dprintf("Can't cope with DTL at logical 0\n");
420         return H_HARDWARE;
421     }
422 
423     size = ldl_be_phys(CPU(cpu)->as, addr + 0x4);
424 
425     if (size < 48) {
426         return H_PARAMETER;
427     }
428 
429     if (!spapr_cpu->vpa_addr) {
430         return H_RESOURCE;
431     }
432 
433     spapr_cpu->dtl_addr = addr;
434     spapr_cpu->dtl_size = size;
435 
436     return H_SUCCESS;
437 }
438 
439 static target_ulong deregister_dtl(PowerPCCPU *cpu, target_ulong addr)
440 {
441     SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
442 
443     spapr_cpu->dtl_addr = 0;
444     spapr_cpu->dtl_size = 0;
445 
446     return H_SUCCESS;
447 }
448 
449 static target_ulong h_register_vpa(PowerPCCPU *cpu, SpaprMachineState *spapr,
450                                    target_ulong opcode, target_ulong *args)
451 {
452     target_ulong flags = args[0];
453     target_ulong procno = args[1];
454     target_ulong vpa = args[2];
455     target_ulong ret = H_PARAMETER;
456     PowerPCCPU *tcpu;
457 
458     tcpu = spapr_find_cpu(procno);
459     if (!tcpu) {
460         return H_PARAMETER;
461     }
462 
463     switch (flags) {
464     case FLAGS_REGISTER_VPA:
465         ret = register_vpa(tcpu, vpa);
466         break;
467 
468     case FLAGS_DEREGISTER_VPA:
469         ret = deregister_vpa(tcpu, vpa);
470         break;
471 
472     case FLAGS_REGISTER_SLBSHADOW:
473         ret = register_slb_shadow(tcpu, vpa);
474         break;
475 
476     case FLAGS_DEREGISTER_SLBSHADOW:
477         ret = deregister_slb_shadow(tcpu, vpa);
478         break;
479 
480     case FLAGS_REGISTER_DTL:
481         ret = register_dtl(tcpu, vpa);
482         break;
483 
484     case FLAGS_DEREGISTER_DTL:
485         ret = deregister_dtl(tcpu, vpa);
486         break;
487     }
488 
489     return ret;
490 }
491 
492 static target_ulong h_cede(PowerPCCPU *cpu, SpaprMachineState *spapr,
493                            target_ulong opcode, target_ulong *args)
494 {
495     CPUPPCState *env = &cpu->env;
496     CPUState *cs = CPU(cpu);
497     SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
498 
499     env->msr |= (1ULL << MSR_EE);
500     hreg_compute_hflags(env);
501     ppc_maybe_interrupt(env);
502 
503     if (spapr_cpu->prod) {
504         spapr_cpu->prod = false;
505         return H_SUCCESS;
506     }
507 
508     if (!cpu_has_work(cs)) {
509         cs->halted = 1;
510         cs->exception_index = EXCP_HLT;
511         cs->exit_request = 1;
512         ppc_maybe_interrupt(env);
513     }
514 
515     return H_SUCCESS;
516 }
517 
518 /*
519  * Confer to self, aka join. Cede could use the same pattern as well, if
520  * EXCP_HLT can be changed to ECXP_HALTED.
521  */
522 static target_ulong h_confer_self(PowerPCCPU *cpu)
523 {
524     CPUState *cs = CPU(cpu);
525     SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
526 
527     if (spapr_cpu->prod) {
528         spapr_cpu->prod = false;
529         return H_SUCCESS;
530     }
531     cs->halted = 1;
532     cs->exception_index = EXCP_HALTED;
533     cs->exit_request = 1;
534     ppc_maybe_interrupt(&cpu->env);
535 
536     return H_SUCCESS;
537 }
538 
539 static target_ulong h_join(PowerPCCPU *cpu, SpaprMachineState *spapr,
540                            target_ulong opcode, target_ulong *args)
541 {
542     CPUPPCState *env = &cpu->env;
543     CPUState *cs;
544     bool last_unjoined = true;
545 
546     if (env->msr & (1ULL << MSR_EE)) {
547         return H_BAD_MODE;
548     }
549 
550     /*
551      * Must not join the last CPU running. Interestingly, no such restriction
552      * for H_CONFER-to-self, but that is probably not intended to be used
553      * when H_JOIN is available.
554      */
555     CPU_FOREACH(cs) {
556         PowerPCCPU *c = POWERPC_CPU(cs);
557         CPUPPCState *e = &c->env;
558         if (c == cpu) {
559             continue;
560         }
561 
562         /* Don't have a way to indicate joined, so use halted && MSR[EE]=0 */
563         if (!cs->halted || (e->msr & (1ULL << MSR_EE))) {
564             last_unjoined = false;
565             break;
566         }
567     }
568     if (last_unjoined) {
569         return H_CONTINUE;
570     }
571 
572     return h_confer_self(cpu);
573 }
574 
575 static target_ulong h_confer(PowerPCCPU *cpu, SpaprMachineState *spapr,
576                            target_ulong opcode, target_ulong *args)
577 {
578     target_long target = args[0];
579     uint32_t dispatch = args[1];
580     CPUState *cs = CPU(cpu);
581     SpaprCpuState *spapr_cpu;
582 
583     assert(tcg_enabled()); /* KVM will have handled this */
584 
585     /*
586      * -1 means confer to all other CPUs without dispatch counter check,
587      *  otherwise it's a targeted confer.
588      */
589     if (target != -1) {
590         PowerPCCPU *target_cpu = spapr_find_cpu(target);
591         uint32_t target_dispatch;
592 
593         if (!target_cpu) {
594             return H_PARAMETER;
595         }
596 
597         /*
598          * target == self is a special case, we wait until prodded, without
599          * dispatch counter check.
600          */
601         if (cpu == target_cpu) {
602             return h_confer_self(cpu);
603         }
604 
605         spapr_cpu = spapr_cpu_state(target_cpu);
606         if (!spapr_cpu->vpa_addr || ((dispatch & 1) == 0)) {
607             return H_SUCCESS;
608         }
609 
610         target_dispatch = ldl_be_phys(cs->as,
611                                   spapr_cpu->vpa_addr + VPA_DISPATCH_COUNTER);
612         if (target_dispatch != dispatch) {
613             return H_SUCCESS;
614         }
615 
616         /*
617          * The targeted confer does not do anything special beyond yielding
618          * the current vCPU, but even this should be better than nothing.
619          * At least for single-threaded tcg, it gives the target a chance to
620          * run before we run again. Multi-threaded tcg does not really do
621          * anything with EXCP_YIELD yet.
622          */
623     }
624 
625     cs->exception_index = EXCP_YIELD;
626     cs->exit_request = 1;
627     cpu_loop_exit(cs);
628 
629     return H_SUCCESS;
630 }
631 
632 static target_ulong h_prod(PowerPCCPU *cpu, SpaprMachineState *spapr,
633                            target_ulong opcode, target_ulong *args)
634 {
635     target_long target = args[0];
636     PowerPCCPU *tcpu;
637     CPUState *cs;
638     SpaprCpuState *spapr_cpu;
639 
640     tcpu = spapr_find_cpu(target);
641     cs = CPU(tcpu);
642     if (!cs) {
643         return H_PARAMETER;
644     }
645 
646     spapr_cpu = spapr_cpu_state(tcpu);
647     spapr_cpu->prod = true;
648     cs->halted = 0;
649     ppc_maybe_interrupt(&cpu->env);
650     qemu_cpu_kick(cs);
651 
652     return H_SUCCESS;
653 }
654 
655 static target_ulong h_rtas(PowerPCCPU *cpu, SpaprMachineState *spapr,
656                            target_ulong opcode, target_ulong *args)
657 {
658     target_ulong rtas_r3 = args[0];
659     uint32_t token = rtas_ld(rtas_r3, 0);
660     uint32_t nargs = rtas_ld(rtas_r3, 1);
661     uint32_t nret = rtas_ld(rtas_r3, 2);
662 
663     return spapr_rtas_call(cpu, spapr, token, nargs, rtas_r3 + 12,
664                            nret, rtas_r3 + 12 + 4*nargs);
665 }
666 
667 static target_ulong h_logical_load(PowerPCCPU *cpu, SpaprMachineState *spapr,
668                                    target_ulong opcode, target_ulong *args)
669 {
670     CPUState *cs = CPU(cpu);
671     target_ulong size = args[0];
672     target_ulong addr = args[1];
673 
674     switch (size) {
675     case 1:
676         args[0] = ldub_phys(cs->as, addr);
677         return H_SUCCESS;
678     case 2:
679         args[0] = lduw_phys(cs->as, addr);
680         return H_SUCCESS;
681     case 4:
682         args[0] = ldl_phys(cs->as, addr);
683         return H_SUCCESS;
684     case 8:
685         args[0] = ldq_phys(cs->as, addr);
686         return H_SUCCESS;
687     }
688     return H_PARAMETER;
689 }
690 
691 static target_ulong h_logical_store(PowerPCCPU *cpu, SpaprMachineState *spapr,
692                                     target_ulong opcode, target_ulong *args)
693 {
694     CPUState *cs = CPU(cpu);
695 
696     target_ulong size = args[0];
697     target_ulong addr = args[1];
698     target_ulong val  = args[2];
699 
700     switch (size) {
701     case 1:
702         stb_phys(cs->as, addr, val);
703         return H_SUCCESS;
704     case 2:
705         stw_phys(cs->as, addr, val);
706         return H_SUCCESS;
707     case 4:
708         stl_phys(cs->as, addr, val);
709         return H_SUCCESS;
710     case 8:
711         stq_phys(cs->as, addr, val);
712         return H_SUCCESS;
713     }
714     return H_PARAMETER;
715 }
716 
717 static target_ulong h_logical_memop(PowerPCCPU *cpu, SpaprMachineState *spapr,
718                                     target_ulong opcode, target_ulong *args)
719 {
720     CPUState *cs = CPU(cpu);
721 
722     target_ulong dst   = args[0]; /* Destination address */
723     target_ulong src   = args[1]; /* Source address */
724     target_ulong esize = args[2]; /* Element size (0=1,1=2,2=4,3=8) */
725     target_ulong count = args[3]; /* Element count */
726     target_ulong op    = args[4]; /* 0 = copy, 1 = invert */
727     uint64_t tmp;
728     unsigned int mask = (1 << esize) - 1;
729     int step = 1 << esize;
730 
731     if (count > 0x80000000) {
732         return H_PARAMETER;
733     }
734 
735     if ((dst & mask) || (src & mask) || (op > 1)) {
736         return H_PARAMETER;
737     }
738 
739     if (dst >= src && dst < (src + (count << esize))) {
740             dst = dst + ((count - 1) << esize);
741             src = src + ((count - 1) << esize);
742             step = -step;
743     }
744 
745     while (count--) {
746         switch (esize) {
747         case 0:
748             tmp = ldub_phys(cs->as, src);
749             break;
750         case 1:
751             tmp = lduw_phys(cs->as, src);
752             break;
753         case 2:
754             tmp = ldl_phys(cs->as, src);
755             break;
756         case 3:
757             tmp = ldq_phys(cs->as, src);
758             break;
759         default:
760             return H_PARAMETER;
761         }
762         if (op == 1) {
763             tmp = ~tmp;
764         }
765         switch (esize) {
766         case 0:
767             stb_phys(cs->as, dst, tmp);
768             break;
769         case 1:
770             stw_phys(cs->as, dst, tmp);
771             break;
772         case 2:
773             stl_phys(cs->as, dst, tmp);
774             break;
775         case 3:
776             stq_phys(cs->as, dst, tmp);
777             break;
778         }
779         dst = dst + step;
780         src = src + step;
781     }
782 
783     return H_SUCCESS;
784 }
785 
786 static target_ulong h_logical_icbi(PowerPCCPU *cpu, SpaprMachineState *spapr,
787                                    target_ulong opcode, target_ulong *args)
788 {
789     /* Nothing to do on emulation, KVM will trap this in the kernel */
790     return H_SUCCESS;
791 }
792 
793 static target_ulong h_logical_dcbf(PowerPCCPU *cpu, SpaprMachineState *spapr,
794                                    target_ulong opcode, target_ulong *args)
795 {
796     /* Nothing to do on emulation, KVM will trap this in the kernel */
797     return H_SUCCESS;
798 }
799 
800 static target_ulong h_set_mode_resource_set_ciabr(PowerPCCPU *cpu,
801                                                   SpaprMachineState *spapr,
802                                                   target_ulong mflags,
803                                                   target_ulong value1,
804                                                   target_ulong value2)
805 {
806     CPUPPCState *env = &cpu->env;
807 
808     assert(tcg_enabled()); /* KVM will have handled this */
809 
810     if (mflags) {
811         return H_UNSUPPORTED_FLAG;
812     }
813     if (value2) {
814         return H_P4;
815     }
816     if ((value1 & PPC_BITMASK(62, 63)) == 0x3) {
817         return H_P3;
818     }
819 
820     ppc_store_ciabr(env, value1);
821 
822     return H_SUCCESS;
823 }
824 
825 static target_ulong h_set_mode_resource_set_dawr0(PowerPCCPU *cpu,
826                                                   SpaprMachineState *spapr,
827                                                   target_ulong mflags,
828                                                   target_ulong value1,
829                                                   target_ulong value2)
830 {
831     CPUPPCState *env = &cpu->env;
832 
833     assert(tcg_enabled()); /* KVM will have handled this */
834 
835     if (mflags) {
836         return H_UNSUPPORTED_FLAG;
837     }
838     if (value2 & PPC_BIT(61)) {
839         return H_P4;
840     }
841 
842     ppc_store_dawr0(env, value1);
843     ppc_store_dawrx0(env, value2);
844 
845     return H_SUCCESS;
846 }
847 
848 static target_ulong h_set_mode_resource_le(PowerPCCPU *cpu,
849                                            SpaprMachineState *spapr,
850                                            target_ulong mflags,
851                                            target_ulong value1,
852                                            target_ulong value2)
853 {
854     if (value1) {
855         return H_P3;
856     }
857     if (value2) {
858         return H_P4;
859     }
860 
861     switch (mflags) {
862     case H_SET_MODE_ENDIAN_BIG:
863         spapr_set_all_lpcrs(0, LPCR_ILE);
864         spapr_pci_switch_vga(spapr, true);
865         return H_SUCCESS;
866 
867     case H_SET_MODE_ENDIAN_LITTLE:
868         spapr_set_all_lpcrs(LPCR_ILE, LPCR_ILE);
869         spapr_pci_switch_vga(spapr, false);
870         return H_SUCCESS;
871     }
872 
873     return H_UNSUPPORTED_FLAG;
874 }
875 
876 static target_ulong h_set_mode_resource_addr_trans_mode(PowerPCCPU *cpu,
877                                                         SpaprMachineState *spapr,
878                                                         target_ulong mflags,
879                                                         target_ulong value1,
880                                                         target_ulong value2)
881 {
882     if (value1) {
883         return H_P3;
884     }
885 
886     if (value2) {
887         return H_P4;
888     }
889 
890     /*
891      * AIL-1 is not architected, and AIL-2 is not supported by QEMU spapr.
892      * It is supported for faithful emulation of bare metal systems, but for
893      * compatibility concerns we leave it out of the pseries machine.
894      */
895     if (mflags != 0 && mflags != 3) {
896         return H_UNSUPPORTED_FLAG;
897     }
898 
899     if (mflags == 3) {
900         if (!spapr_get_cap(spapr, SPAPR_CAP_AIL_MODE_3)) {
901             return H_UNSUPPORTED_FLAG;
902         }
903     }
904 
905     spapr_set_all_lpcrs(mflags << LPCR_AIL_SHIFT, LPCR_AIL);
906 
907     return H_SUCCESS;
908 }
909 
910 static target_ulong h_set_mode(PowerPCCPU *cpu, SpaprMachineState *spapr,
911                                target_ulong opcode, target_ulong *args)
912 {
913     target_ulong resource = args[1];
914     target_ulong ret = H_P2;
915 
916     switch (resource) {
917     case H_SET_MODE_RESOURCE_SET_CIABR:
918         ret = h_set_mode_resource_set_ciabr(cpu, spapr, args[0], args[2],
919                                             args[3]);
920         break;
921     case H_SET_MODE_RESOURCE_SET_DAWR0:
922         ret = h_set_mode_resource_set_dawr0(cpu, spapr, args[0], args[2],
923                                             args[3]);
924         break;
925     case H_SET_MODE_RESOURCE_LE:
926         ret = h_set_mode_resource_le(cpu, spapr, args[0], args[2], args[3]);
927         break;
928     case H_SET_MODE_RESOURCE_ADDR_TRANS_MODE:
929         ret = h_set_mode_resource_addr_trans_mode(cpu, spapr, args[0],
930                                                   args[2], args[3]);
931         break;
932     }
933 
934     return ret;
935 }
936 
937 static target_ulong h_clean_slb(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_CLEAN_SLB)");
942     return H_FUNCTION;
943 }
944 
945 static target_ulong h_invalidate_pid(PowerPCCPU *cpu, SpaprMachineState *spapr,
946                                      target_ulong opcode, target_ulong *args)
947 {
948     qemu_log_mask(LOG_UNIMP, "Unimplemented SPAPR hcall 0x"TARGET_FMT_lx"%s\n",
949                   opcode, " (H_INVALIDATE_PID)");
950     return H_FUNCTION;
951 }
952 
953 static void spapr_check_setup_free_hpt(SpaprMachineState *spapr,
954                                        uint64_t patbe_old, uint64_t patbe_new)
955 {
956     /*
957      * We have 4 Options:
958      * HASH->HASH || RADIX->RADIX || NOTHING->RADIX : Do Nothing
959      * HASH->RADIX                                  : Free HPT
960      * RADIX->HASH                                  : Allocate HPT
961      * NOTHING->HASH                                : Allocate HPT
962      * Note: NOTHING implies the case where we said the guest could choose
963      *       later and so assumed radix and now it's called H_REG_PROC_TBL
964      */
965 
966     if ((patbe_old & PATE1_GR) == (patbe_new & PATE1_GR)) {
967         /* We assume RADIX, so this catches all the "Do Nothing" cases */
968     } else if (!(patbe_old & PATE1_GR)) {
969         /* HASH->RADIX : Free HPT */
970         spapr_free_hpt(spapr);
971     } else if (!(patbe_new & PATE1_GR)) {
972         /* RADIX->HASH || NOTHING->HASH : Allocate HPT */
973         spapr_setup_hpt(spapr);
974     }
975     return;
976 }
977 
978 #define FLAGS_MASK              0x01FULL
979 #define FLAG_MODIFY             0x10
980 #define FLAG_REGISTER           0x08
981 #define FLAG_RADIX              0x04
982 #define FLAG_HASH_PROC_TBL      0x02
983 #define FLAG_GTSE               0x01
984 
985 static target_ulong h_register_process_table(PowerPCCPU *cpu,
986                                              SpaprMachineState *spapr,
987                                              target_ulong opcode,
988                                              target_ulong *args)
989 {
990     target_ulong flags = args[0];
991     target_ulong proc_tbl = args[1];
992     target_ulong page_size = args[2];
993     target_ulong table_size = args[3];
994     target_ulong update_lpcr = 0;
995     target_ulong table_byte_size;
996     uint64_t cproc;
997 
998     if (flags & ~FLAGS_MASK) { /* Check no reserved bits are set */
999         return H_PARAMETER;
1000     }
1001     if (flags & FLAG_MODIFY) {
1002         if (flags & FLAG_REGISTER) {
1003             /* Check process table alignment */
1004             table_byte_size = 1ULL << (table_size + 12);
1005             if (proc_tbl & (table_byte_size - 1)) {
1006                 qemu_log_mask(LOG_GUEST_ERROR,
1007                     "%s: process table not properly aligned: proc_tbl 0x"
1008                     TARGET_FMT_lx" proc_tbl_size 0x"TARGET_FMT_lx"\n",
1009                     __func__, proc_tbl, table_byte_size);
1010             }
1011             if (flags & FLAG_RADIX) { /* Register new RADIX process table */
1012                 if (proc_tbl & 0xfff || proc_tbl >> 60) {
1013                     return H_P2;
1014                 } else if (page_size) {
1015                     return H_P3;
1016                 } else if (table_size > 24) {
1017                     return H_P4;
1018                 }
1019                 cproc = PATE1_GR | proc_tbl | table_size;
1020             } else { /* Register new HPT process table */
1021                 if (flags & FLAG_HASH_PROC_TBL) { /* Hash with Segment Tables */
1022                     /* TODO - Not Supported */
1023                     /* Technically caused by flag bits => H_PARAMETER */
1024                     return H_PARAMETER;
1025                 } else { /* Hash with SLB */
1026                     if (proc_tbl >> 38) {
1027                         return H_P2;
1028                     } else if (page_size & ~0x7) {
1029                         return H_P3;
1030                     } else if (table_size > 24) {
1031                         return H_P4;
1032                     }
1033                 }
1034                 cproc = (proc_tbl << 25) | page_size << 5 | table_size;
1035             }
1036 
1037         } else { /* Deregister current process table */
1038             /*
1039              * Set to benign value: (current GR) | 0. This allows
1040              * deregistration in KVM to succeed even if the radix bit
1041              * in flags doesn't match the radix bit in the old PATE.
1042              */
1043             cproc = spapr->patb_entry & PATE1_GR;
1044         }
1045     } else { /* Maintain current registration */
1046         if (!(flags & FLAG_RADIX) != !(spapr->patb_entry & PATE1_GR)) {
1047             /* Technically caused by flag bits => H_PARAMETER */
1048             return H_PARAMETER; /* Existing Process Table Mismatch */
1049         }
1050         cproc = spapr->patb_entry;
1051     }
1052 
1053     /* Check if we need to setup OR free the hpt */
1054     spapr_check_setup_free_hpt(spapr, spapr->patb_entry, cproc);
1055 
1056     spapr->patb_entry = cproc; /* Save new process table */
1057 
1058     /* Update the UPRT, HR and GTSE bits in the LPCR for all cpus */
1059     if (flags & FLAG_RADIX)     /* Radix must use process tables, also set HR */
1060         update_lpcr |= (LPCR_UPRT | LPCR_HR);
1061     else if (flags & FLAG_HASH_PROC_TBL) /* Hash with process tables */
1062         update_lpcr |= LPCR_UPRT;
1063     if (flags & FLAG_GTSE)      /* Guest translation shootdown enable */
1064         update_lpcr |= LPCR_GTSE;
1065 
1066     spapr_set_all_lpcrs(update_lpcr, LPCR_UPRT | LPCR_HR | LPCR_GTSE);
1067 
1068     if (kvm_enabled()) {
1069         return kvmppc_configure_v3_mmu(cpu, flags & FLAG_RADIX,
1070                                        flags & FLAG_GTSE, cproc);
1071     }
1072     return H_SUCCESS;
1073 }
1074 
1075 #define H_SIGNAL_SYS_RESET_ALL         -1
1076 #define H_SIGNAL_SYS_RESET_ALLBUTSELF  -2
1077 
1078 static target_ulong h_signal_sys_reset(PowerPCCPU *cpu,
1079                                        SpaprMachineState *spapr,
1080                                        target_ulong opcode, target_ulong *args)
1081 {
1082     target_long target = args[0];
1083     CPUState *cs;
1084 
1085     if (target < 0) {
1086         /* Broadcast */
1087         if (target < H_SIGNAL_SYS_RESET_ALLBUTSELF) {
1088             return H_PARAMETER;
1089         }
1090 
1091         CPU_FOREACH(cs) {
1092             PowerPCCPU *c = POWERPC_CPU(cs);
1093 
1094             if (target == H_SIGNAL_SYS_RESET_ALLBUTSELF) {
1095                 if (c == cpu) {
1096                     continue;
1097                 }
1098             }
1099             run_on_cpu(cs, spapr_do_system_reset_on_cpu, RUN_ON_CPU_NULL);
1100         }
1101         return H_SUCCESS;
1102 
1103     } else {
1104         /* Unicast */
1105         cs = CPU(spapr_find_cpu(target));
1106         if (cs) {
1107             run_on_cpu(cs, spapr_do_system_reset_on_cpu, RUN_ON_CPU_NULL);
1108             return H_SUCCESS;
1109         }
1110         return H_PARAMETER;
1111     }
1112 }
1113 
1114 /* Returns either a logical PVR or zero if none was found */
1115 static uint32_t cas_check_pvr(PowerPCCPU *cpu, uint32_t max_compat,
1116                               target_ulong *addr, bool *raw_mode_supported)
1117 {
1118     bool explicit_match = false; /* Matched the CPU's real PVR */
1119     uint32_t best_compat = 0;
1120     int i;
1121 
1122     /*
1123      * We scan the supplied table of PVRs looking for two things
1124      *   1. Is our real CPU PVR in the list?
1125      *   2. What's the "best" listed logical PVR
1126      */
1127     for (i = 0; i < 512; ++i) {
1128         uint32_t pvr, pvr_mask;
1129 
1130         pvr_mask = ldl_be_phys(&address_space_memory, *addr);
1131         pvr = ldl_be_phys(&address_space_memory, *addr + 4);
1132         *addr += 8;
1133 
1134         if (~pvr_mask & pvr) {
1135             break; /* Terminator record */
1136         }
1137 
1138         if ((cpu->env.spr[SPR_PVR] & pvr_mask) == (pvr & pvr_mask)) {
1139             explicit_match = true;
1140         } else {
1141             if (ppc_check_compat(cpu, pvr, best_compat, max_compat)) {
1142                 best_compat = pvr;
1143             }
1144         }
1145     }
1146 
1147     *raw_mode_supported = explicit_match;
1148 
1149     /* Parsing finished */
1150     trace_spapr_cas_pvr(cpu->compat_pvr, explicit_match, best_compat);
1151 
1152     return best_compat;
1153 }
1154 
1155 static
1156 target_ulong do_client_architecture_support(PowerPCCPU *cpu,
1157                                             SpaprMachineState *spapr,
1158                                             target_ulong vec,
1159                                             target_ulong fdt_bufsize)
1160 {
1161     target_ulong ov_table; /* Working address in data buffer */
1162     uint32_t cas_pvr;
1163     SpaprOptionVector *ov1_guest, *ov5_guest;
1164     bool guest_radix;
1165     bool raw_mode_supported = false;
1166     bool guest_xive;
1167     CPUState *cs;
1168     void *fdt;
1169     uint32_t max_compat = spapr->max_compat_pvr;
1170 
1171     /* CAS is supposed to be called early when only the boot vCPU is active. */
1172     CPU_FOREACH(cs) {
1173         if (cs == CPU(cpu)) {
1174             continue;
1175         }
1176         if (!cs->halted) {
1177             warn_report("guest has multiple active vCPUs at CAS, which is not allowed");
1178             return H_MULTI_THREADS_ACTIVE;
1179         }
1180     }
1181 
1182     cas_pvr = cas_check_pvr(cpu, max_compat, &vec, &raw_mode_supported);
1183     if (!cas_pvr && (!raw_mode_supported || max_compat)) {
1184         /*
1185          * We couldn't find a suitable compatibility mode, and either
1186          * the guest doesn't support "raw" mode for this CPU, or "raw"
1187          * mode is disabled because a maximum compat mode is set.
1188          */
1189         error_report("Couldn't negotiate a suitable PVR during CAS");
1190         return H_HARDWARE;
1191     }
1192 
1193     /* Update CPUs */
1194     if (cpu->compat_pvr != cas_pvr) {
1195         Error *local_err = NULL;
1196 
1197         if (ppc_set_compat_all(cas_pvr, &local_err) < 0) {
1198             /* We fail to set compat mode (likely because running with KVM PR),
1199              * but maybe we can fallback to raw mode if the guest supports it.
1200              */
1201             if (!raw_mode_supported) {
1202                 error_report_err(local_err);
1203                 return H_HARDWARE;
1204             }
1205             error_free(local_err);
1206         }
1207     }
1208 
1209     /* For the future use: here @ov_table points to the first option vector */
1210     ov_table = vec;
1211 
1212     ov1_guest = spapr_ovec_parse_vector(ov_table, 1);
1213     if (!ov1_guest) {
1214         warn_report("guest didn't provide option vector 1");
1215         return H_PARAMETER;
1216     }
1217     ov5_guest = spapr_ovec_parse_vector(ov_table, 5);
1218     if (!ov5_guest) {
1219         spapr_ovec_cleanup(ov1_guest);
1220         warn_report("guest didn't provide option vector 5");
1221         return H_PARAMETER;
1222     }
1223     if (spapr_ovec_test(ov5_guest, OV5_MMU_BOTH)) {
1224         error_report("guest requested hash and radix MMU, which is invalid.");
1225         exit(EXIT_FAILURE);
1226     }
1227     if (spapr_ovec_test(ov5_guest, OV5_XIVE_BOTH)) {
1228         error_report("guest requested an invalid interrupt mode");
1229         exit(EXIT_FAILURE);
1230     }
1231 
1232     guest_radix = spapr_ovec_test(ov5_guest, OV5_MMU_RADIX_300);
1233 
1234     guest_xive = spapr_ovec_test(ov5_guest, OV5_XIVE_EXPLOIT);
1235 
1236     /*
1237      * HPT resizing is a bit of a special case, because when enabled
1238      * we assume an HPT guest will support it until it says it
1239      * doesn't, instead of assuming it won't support it until it says
1240      * it does.  Strictly speaking that approach could break for
1241      * guests which don't make a CAS call, but those are so old we
1242      * don't care about them.  Without that assumption we'd have to
1243      * make at least a temporary allocation of an HPT sized for max
1244      * memory, which could be impossibly difficult under KVM HV if
1245      * maxram is large.
1246      */
1247     if (!guest_radix && !spapr_ovec_test(ov5_guest, OV5_HPT_RESIZE)) {
1248         int maxshift = spapr_hpt_shift_for_ramsize(MACHINE(spapr)->maxram_size);
1249 
1250         if (spapr->resize_hpt == SPAPR_RESIZE_HPT_REQUIRED) {
1251             error_report(
1252                 "h_client_architecture_support: Guest doesn't support HPT resizing, but resize-hpt=required");
1253             exit(1);
1254         }
1255 
1256         if (spapr->htab_shift < maxshift) {
1257             /* Guest doesn't know about HPT resizing, so we
1258              * pre-emptively resize for the maximum permitted RAM.  At
1259              * the point this is called, nothing should have been
1260              * entered into the existing HPT */
1261             spapr_reallocate_hpt(spapr, maxshift, &error_fatal);
1262             push_sregs_to_kvm_pr(spapr);
1263         }
1264     }
1265 
1266     /* NOTE: there are actually a number of ov5 bits where input from the
1267      * guest is always zero, and the platform/QEMU enables them independently
1268      * of guest input. To model these properly we'd want some sort of mask,
1269      * but since they only currently apply to memory migration as defined
1270      * by LoPAPR 1.1, 14.5.4.8, which QEMU doesn't implement, we don't need
1271      * to worry about this for now.
1272      */
1273 
1274     /* full range of negotiated ov5 capabilities */
1275     spapr_ovec_intersect(spapr->ov5_cas, spapr->ov5, ov5_guest);
1276     spapr_ovec_cleanup(ov5_guest);
1277 
1278     spapr_check_mmu_mode(guest_radix);
1279 
1280     spapr->cas_pre_isa3_guest = !spapr_ovec_test(ov1_guest, OV1_PPC_3_00);
1281     spapr_ovec_cleanup(ov1_guest);
1282 
1283     /*
1284      * Check for NUMA affinity conditions now that we know which NUMA
1285      * affinity the guest will use.
1286      */
1287     spapr_numa_associativity_check(spapr);
1288 
1289     /*
1290      * Ensure the guest asks for an interrupt mode we support;
1291      * otherwise terminate the boot.
1292      */
1293     if (guest_xive) {
1294         if (!spapr->irq->xive) {
1295             error_report(
1296 "Guest requested unavailable interrupt mode (XIVE), try the ic-mode=xive or ic-mode=dual machine property");
1297             exit(EXIT_FAILURE);
1298         }
1299     } else {
1300         if (!spapr->irq->xics) {
1301             error_report(
1302 "Guest requested unavailable interrupt mode (XICS), either don't set the ic-mode machine property or try ic-mode=xics or ic-mode=dual");
1303             exit(EXIT_FAILURE);
1304         }
1305     }
1306 
1307     spapr_irq_update_active_intc(spapr);
1308 
1309     /*
1310      * Process all pending hot-plug/unplug requests now. An updated full
1311      * rendered FDT will be returned to the guest.
1312      */
1313     spapr_drc_reset_all(spapr);
1314     spapr_clear_pending_hotplug_events(spapr);
1315 
1316     /*
1317      * If spapr_machine_reset() did not set up a HPT but one is necessary
1318      * (because the guest isn't going to use radix) then set it up here.
1319      */
1320     if ((spapr->patb_entry & PATE1_GR) && !guest_radix) {
1321         /* legacy hash or new hash: */
1322         spapr_setup_hpt(spapr);
1323     }
1324 
1325     fdt = spapr_build_fdt(spapr, spapr->vof != NULL, fdt_bufsize);
1326     g_free(spapr->fdt_blob);
1327     spapr->fdt_size = fdt_totalsize(fdt);
1328     spapr->fdt_initial_size = spapr->fdt_size;
1329     spapr->fdt_blob = fdt;
1330 
1331     /*
1332      * Set the machine->fdt pointer again since we just freed
1333      * it above (by freeing spapr->fdt_blob). We set this
1334      * pointer to enable support for the 'dumpdtb' QMP/HMP
1335      * command.
1336      */
1337     MACHINE(spapr)->fdt = fdt;
1338 
1339     return H_SUCCESS;
1340 }
1341 
1342 static target_ulong h_client_architecture_support(PowerPCCPU *cpu,
1343                                                   SpaprMachineState *spapr,
1344                                                   target_ulong opcode,
1345                                                   target_ulong *args)
1346 {
1347     target_ulong vec = ppc64_phys_to_real(args[0]);
1348     target_ulong fdt_buf = args[1];
1349     target_ulong fdt_bufsize = args[2];
1350     target_ulong ret;
1351     SpaprDeviceTreeUpdateHeader hdr = { .version_id = 1 };
1352 
1353     if (fdt_bufsize < sizeof(hdr)) {
1354         error_report("SLOF provided insufficient CAS buffer "
1355                      TARGET_FMT_lu " (min: %zu)", fdt_bufsize, sizeof(hdr));
1356         exit(EXIT_FAILURE);
1357     }
1358 
1359     fdt_bufsize -= sizeof(hdr);
1360 
1361     ret = do_client_architecture_support(cpu, spapr, vec, fdt_bufsize);
1362     if (ret == H_SUCCESS) {
1363         _FDT((fdt_pack(spapr->fdt_blob)));
1364         spapr->fdt_size = fdt_totalsize(spapr->fdt_blob);
1365         spapr->fdt_initial_size = spapr->fdt_size;
1366 
1367         cpu_physical_memory_write(fdt_buf, &hdr, sizeof(hdr));
1368         cpu_physical_memory_write(fdt_buf + sizeof(hdr), spapr->fdt_blob,
1369                                   spapr->fdt_size);
1370         trace_spapr_cas_continue(spapr->fdt_size + sizeof(hdr));
1371     }
1372 
1373     return ret;
1374 }
1375 
1376 target_ulong spapr_vof_client_architecture_support(MachineState *ms,
1377                                                    CPUState *cs,
1378                                                    target_ulong ovec_addr)
1379 {
1380     SpaprMachineState *spapr = SPAPR_MACHINE(ms);
1381 
1382     target_ulong ret = do_client_architecture_support(POWERPC_CPU(cs), spapr,
1383                                                       ovec_addr, FDT_MAX_SIZE);
1384 
1385     /*
1386      * This adds stdout and generates phandles for boottime and CAS FDTs.
1387      * It is alright to update the FDT here as do_client_architecture_support()
1388      * does not pack it.
1389      */
1390     spapr_vof_client_dt_finalize(spapr, spapr->fdt_blob);
1391 
1392     return ret;
1393 }
1394 
1395 static target_ulong h_get_cpu_characteristics(PowerPCCPU *cpu,
1396                                               SpaprMachineState *spapr,
1397                                               target_ulong opcode,
1398                                               target_ulong *args)
1399 {
1400     uint64_t characteristics = H_CPU_CHAR_HON_BRANCH_HINTS &
1401                                ~H_CPU_CHAR_THR_RECONF_TRIG;
1402     uint64_t behaviour = H_CPU_BEHAV_FAVOUR_SECURITY;
1403     uint8_t safe_cache = spapr_get_cap(spapr, SPAPR_CAP_CFPC);
1404     uint8_t safe_bounds_check = spapr_get_cap(spapr, SPAPR_CAP_SBBC);
1405     uint8_t safe_indirect_branch = spapr_get_cap(spapr, SPAPR_CAP_IBS);
1406     uint8_t count_cache_flush_assist = spapr_get_cap(spapr,
1407                                                      SPAPR_CAP_CCF_ASSIST);
1408 
1409     switch (safe_cache) {
1410     case SPAPR_CAP_WORKAROUND:
1411         characteristics |= H_CPU_CHAR_L1D_FLUSH_ORI30;
1412         characteristics |= H_CPU_CHAR_L1D_FLUSH_TRIG2;
1413         characteristics |= H_CPU_CHAR_L1D_THREAD_PRIV;
1414         behaviour |= H_CPU_BEHAV_L1D_FLUSH_PR;
1415         break;
1416     case SPAPR_CAP_FIXED:
1417         behaviour |= H_CPU_BEHAV_NO_L1D_FLUSH_ENTRY;
1418         behaviour |= H_CPU_BEHAV_NO_L1D_FLUSH_UACCESS;
1419         break;
1420     default: /* broken */
1421         assert(safe_cache == SPAPR_CAP_BROKEN);
1422         behaviour |= H_CPU_BEHAV_L1D_FLUSH_PR;
1423         break;
1424     }
1425 
1426     switch (safe_bounds_check) {
1427     case SPAPR_CAP_WORKAROUND:
1428         characteristics |= H_CPU_CHAR_SPEC_BAR_ORI31;
1429         behaviour |= H_CPU_BEHAV_BNDS_CHK_SPEC_BAR;
1430         break;
1431     case SPAPR_CAP_FIXED:
1432         break;
1433     default: /* broken */
1434         assert(safe_bounds_check == SPAPR_CAP_BROKEN);
1435         behaviour |= H_CPU_BEHAV_BNDS_CHK_SPEC_BAR;
1436         break;
1437     }
1438 
1439     switch (safe_indirect_branch) {
1440     case SPAPR_CAP_FIXED_NA:
1441         break;
1442     case SPAPR_CAP_FIXED_CCD:
1443         characteristics |= H_CPU_CHAR_CACHE_COUNT_DIS;
1444         break;
1445     case SPAPR_CAP_FIXED_IBS:
1446         characteristics |= H_CPU_CHAR_BCCTRL_SERIALISED;
1447         break;
1448     case SPAPR_CAP_WORKAROUND:
1449         behaviour |= H_CPU_BEHAV_FLUSH_COUNT_CACHE;
1450         if (count_cache_flush_assist) {
1451             characteristics |= H_CPU_CHAR_BCCTR_FLUSH_ASSIST;
1452         }
1453         break;
1454     default: /* broken */
1455         assert(safe_indirect_branch == SPAPR_CAP_BROKEN);
1456         break;
1457     }
1458 
1459     args[0] = characteristics;
1460     args[1] = behaviour;
1461     return H_SUCCESS;
1462 }
1463 
1464 static target_ulong h_update_dt(PowerPCCPU *cpu, SpaprMachineState *spapr,
1465                                 target_ulong opcode, target_ulong *args)
1466 {
1467     target_ulong dt = ppc64_phys_to_real(args[0]);
1468     struct fdt_header hdr = { 0 };
1469     unsigned cb;
1470     SpaprMachineClass *smc = SPAPR_MACHINE_GET_CLASS(spapr);
1471     void *fdt;
1472 
1473     cpu_physical_memory_read(dt, &hdr, sizeof(hdr));
1474     cb = fdt32_to_cpu(hdr.totalsize);
1475 
1476     if (!smc->update_dt_enabled) {
1477         return H_SUCCESS;
1478     }
1479 
1480     /* Check that the fdt did not grow out of proportion */
1481     if (cb > spapr->fdt_initial_size * 2) {
1482         trace_spapr_update_dt_failed_size(spapr->fdt_initial_size, cb,
1483                                           fdt32_to_cpu(hdr.magic));
1484         return H_PARAMETER;
1485     }
1486 
1487     fdt = g_malloc0(cb);
1488     cpu_physical_memory_read(dt, fdt, cb);
1489 
1490     /* Check the fdt consistency */
1491     if (fdt_check_full(fdt, cb)) {
1492         trace_spapr_update_dt_failed_check(spapr->fdt_initial_size, cb,
1493                                            fdt32_to_cpu(hdr.magic));
1494         return H_PARAMETER;
1495     }
1496 
1497     g_free(spapr->fdt_blob);
1498     spapr->fdt_size = cb;
1499     spapr->fdt_blob = fdt;
1500     trace_spapr_update_dt(cb);
1501 
1502     return H_SUCCESS;
1503 }
1504 
1505 static spapr_hcall_fn papr_hypercall_table[(MAX_HCALL_OPCODE / 4) + 1];
1506 static spapr_hcall_fn kvmppc_hypercall_table[KVMPPC_HCALL_MAX - KVMPPC_HCALL_BASE + 1];
1507 static spapr_hcall_fn svm_hypercall_table[(SVM_HCALL_MAX - SVM_HCALL_BASE) / 4 + 1];
1508 
1509 void spapr_register_hypercall(target_ulong opcode, spapr_hcall_fn fn)
1510 {
1511     spapr_hcall_fn *slot;
1512 
1513     if (opcode <= MAX_HCALL_OPCODE) {
1514         assert((opcode & 0x3) == 0);
1515 
1516         slot = &papr_hypercall_table[opcode / 4];
1517     } else if (opcode >= SVM_HCALL_BASE && opcode <= SVM_HCALL_MAX) {
1518         /* we only have SVM-related hcall numbers assigned in multiples of 4 */
1519         assert((opcode & 0x3) == 0);
1520 
1521         slot = &svm_hypercall_table[(opcode - SVM_HCALL_BASE) / 4];
1522     } else {
1523         assert((opcode >= KVMPPC_HCALL_BASE) && (opcode <= KVMPPC_HCALL_MAX));
1524 
1525         slot = &kvmppc_hypercall_table[opcode - KVMPPC_HCALL_BASE];
1526     }
1527 
1528     assert(!(*slot));
1529     *slot = fn;
1530 }
1531 
1532 void spapr_unregister_hypercall(target_ulong opcode)
1533 {
1534     spapr_hcall_fn *slot;
1535 
1536     if (opcode <= MAX_HCALL_OPCODE) {
1537         assert((opcode & 0x3) == 0);
1538 
1539         slot = &papr_hypercall_table[opcode / 4];
1540     } else if (opcode >= SVM_HCALL_BASE && opcode <= SVM_HCALL_MAX) {
1541         /* we only have SVM-related hcall numbers assigned in multiples of 4 */
1542         assert((opcode & 0x3) == 0);
1543 
1544         slot = &svm_hypercall_table[(opcode - SVM_HCALL_BASE) / 4];
1545     } else {
1546         assert((opcode >= KVMPPC_HCALL_BASE) && (opcode <= KVMPPC_HCALL_MAX));
1547 
1548         slot = &kvmppc_hypercall_table[opcode - KVMPPC_HCALL_BASE];
1549     }
1550 
1551     *slot = NULL;
1552 }
1553 
1554 target_ulong spapr_hypercall(PowerPCCPU *cpu, target_ulong opcode,
1555                              target_ulong *args)
1556 {
1557     SpaprMachineState *spapr = SPAPR_MACHINE(qdev_get_machine());
1558 
1559     if ((opcode <= MAX_HCALL_OPCODE)
1560         && ((opcode & 0x3) == 0)) {
1561         spapr_hcall_fn fn = papr_hypercall_table[opcode / 4];
1562 
1563         if (fn) {
1564             return fn(cpu, spapr, opcode, args);
1565         }
1566     } else if ((opcode >= SVM_HCALL_BASE) &&
1567                (opcode <= SVM_HCALL_MAX)) {
1568         spapr_hcall_fn fn = svm_hypercall_table[(opcode - SVM_HCALL_BASE) / 4];
1569 
1570         if (fn) {
1571             return fn(cpu, spapr, opcode, args);
1572         }
1573     } else if ((opcode >= KVMPPC_HCALL_BASE) &&
1574                (opcode <= KVMPPC_HCALL_MAX)) {
1575         spapr_hcall_fn fn = kvmppc_hypercall_table[opcode - KVMPPC_HCALL_BASE];
1576 
1577         if (fn) {
1578             return fn(cpu, spapr, opcode, args);
1579         }
1580     }
1581 
1582     qemu_log_mask(LOG_UNIMP, "Unimplemented SPAPR hcall 0x" TARGET_FMT_lx "\n",
1583                   opcode);
1584     return H_FUNCTION;
1585 }
1586 
1587 #ifdef CONFIG_TCG
1588 static void hypercall_register_softmmu(void)
1589 {
1590     /* DO NOTHING */
1591 }
1592 #else
1593 static target_ulong h_softmmu(PowerPCCPU *cpu, SpaprMachineState *spapr,
1594                             target_ulong opcode, target_ulong *args)
1595 {
1596     g_assert_not_reached();
1597 }
1598 
1599 static void hypercall_register_softmmu(void)
1600 {
1601     /* hcall-pft */
1602     spapr_register_hypercall(H_ENTER, h_softmmu);
1603     spapr_register_hypercall(H_REMOVE, h_softmmu);
1604     spapr_register_hypercall(H_PROTECT, h_softmmu);
1605     spapr_register_hypercall(H_READ, h_softmmu);
1606 
1607     /* hcall-bulk */
1608     spapr_register_hypercall(H_BULK_REMOVE, h_softmmu);
1609 }
1610 #endif
1611 
1612 static void hypercall_register_types(void)
1613 {
1614     hypercall_register_softmmu();
1615 
1616     /* hcall-hpt-resize */
1617     spapr_register_hypercall(H_RESIZE_HPT_PREPARE, h_resize_hpt_prepare);
1618     spapr_register_hypercall(H_RESIZE_HPT_COMMIT, h_resize_hpt_commit);
1619 
1620     /* hcall-splpar */
1621     spapr_register_hypercall(H_REGISTER_VPA, h_register_vpa);
1622     spapr_register_hypercall(H_CEDE, h_cede);
1623     spapr_register_hypercall(H_CONFER, h_confer);
1624     spapr_register_hypercall(H_PROD, h_prod);
1625 
1626     /* hcall-join */
1627     spapr_register_hypercall(H_JOIN, h_join);
1628 
1629     spapr_register_hypercall(H_SIGNAL_SYS_RESET, h_signal_sys_reset);
1630 
1631     /* processor register resource access h-calls */
1632     spapr_register_hypercall(H_SET_SPRG0, h_set_sprg0);
1633     spapr_register_hypercall(H_SET_DABR, h_set_dabr);
1634     spapr_register_hypercall(H_SET_XDABR, h_set_xdabr);
1635     spapr_register_hypercall(H_PAGE_INIT, h_page_init);
1636     spapr_register_hypercall(H_SET_MODE, h_set_mode);
1637 
1638     /* In Memory Table MMU h-calls */
1639     spapr_register_hypercall(H_CLEAN_SLB, h_clean_slb);
1640     spapr_register_hypercall(H_INVALIDATE_PID, h_invalidate_pid);
1641     spapr_register_hypercall(H_REGISTER_PROC_TBL, h_register_process_table);
1642 
1643     /* hcall-get-cpu-characteristics */
1644     spapr_register_hypercall(H_GET_CPU_CHARACTERISTICS,
1645                              h_get_cpu_characteristics);
1646 
1647     /* "debugger" hcalls (also used by SLOF). Note: We do -not- differentiate
1648      * here between the "CI" and the "CACHE" variants, they will use whatever
1649      * mapping attributes qemu is using. When using KVM, the kernel will
1650      * enforce the attributes more strongly
1651      */
1652     spapr_register_hypercall(H_LOGICAL_CI_LOAD, h_logical_load);
1653     spapr_register_hypercall(H_LOGICAL_CI_STORE, h_logical_store);
1654     spapr_register_hypercall(H_LOGICAL_CACHE_LOAD, h_logical_load);
1655     spapr_register_hypercall(H_LOGICAL_CACHE_STORE, h_logical_store);
1656     spapr_register_hypercall(H_LOGICAL_ICBI, h_logical_icbi);
1657     spapr_register_hypercall(H_LOGICAL_DCBF, h_logical_dcbf);
1658     spapr_register_hypercall(KVMPPC_H_LOGICAL_MEMOP, h_logical_memop);
1659 
1660     /* qemu/KVM-PPC specific hcalls */
1661     spapr_register_hypercall(KVMPPC_H_RTAS, h_rtas);
1662 
1663     /* ibm,client-architecture-support support */
1664     spapr_register_hypercall(KVMPPC_H_CAS, h_client_architecture_support);
1665 
1666     spapr_register_hypercall(KVMPPC_H_UPDATE_DT, h_update_dt);
1667 }
1668 
1669 type_init(hypercall_register_types)
1670