master
c 1,697 lines 51 KB
Raw
1 #include "qemu/osdep.h"
2 #include "qemu/cutils.h"
3 #include "qapi/error.h"
4 #include "system/hw_accel.h"
5 #include "system/physmem.h"
6 #include "system/runstate.h"
7 #include "system/tcg.h"
8 #include "qemu/log.h"
9 #include "qemu/main-loop.h"
10 #include "qemu/module.h"
11 #include "qemu/error-report.h"
12 #include "exec/translation-block.h"
13 #include "exec/target_page.h"
14 #include "helper_regs.h"
15 #include "hw/ppc/ppc.h"
16 #include "hw/ppc/spapr.h"
17 #include "hw/ppc/spapr_cpu_core.h"
18 #include "hw/ppc/spapr_nested.h"
19 #include "mmu-hash64.h"
20 #include "cpu-models.h"
21 #include "trace.h"
22 #include "kvm_ppc.h"
23 #include "hw/ppc/fdt.h"
24 #include "hw/ppc/spapr_ovec.h"
25 #include "hw/ppc/spapr_numa.h"
26 #include "mmu-book3s-v3.h"
27 #include "hw/mem/memory-device.h"
28 #include "exec/cpu-common.h"
29
30 bool is_ram_address(SpaprMachineState *spapr, hwaddr addr)
31 {
32 MachineState *machine = MACHINE(spapr);
33 DeviceMemoryState *dms = machine->device_memory;
34
35 if (addr < machine->ram_size) {
36 return true;
37 }
38 if (dms && (addr >= dms->base)
39 && ((addr - dms->base) < memory_region_size(&dms->mr))) {
40 return true;
41 }
42
43 return false;
44 }
45
46 /* Convert a return code from the KVM ioctl()s implementing resize HPT
47 * into a PAPR hypercall return code */
48 static target_ulong resize_hpt_convert_rc(int ret)
49 {
50 if (ret >= 100000) {
51 return H_LONG_BUSY_ORDER_100_SEC;
52 } else if (ret >= 10000) {
53 return H_LONG_BUSY_ORDER_10_SEC;
54 } else if (ret >= 1000) {
55 return H_LONG_BUSY_ORDER_1_SEC;
56 } else if (ret >= 100) {
57 return H_LONG_BUSY_ORDER_100_MSEC;
58 } else if (ret >= 10) {
59 return H_LONG_BUSY_ORDER_10_MSEC;
60 } else if (ret > 0) {
61 return H_LONG_BUSY_ORDER_1_MSEC;
62 }
63
64 switch (ret) {
65 case 0:
66 return H_SUCCESS;
67 case -EPERM:
68 return H_AUTHORITY;
69 case -EINVAL:
70 return H_PARAMETER;
71 case -ENXIO:
72 return H_CLOSED;
73 case -ENOSPC:
74 return H_PTEG_FULL;
75 case -EBUSY:
76 return H_BUSY;
77 case -ENOMEM:
78 return H_NO_MEM;
79 default:
80 return H_HARDWARE;
81 }
82 }
83
84 static target_ulong h_resize_hpt_prepare(PowerPCCPU *cpu,
85 SpaprMachineState *spapr,
86 target_ulong opcode,
87 target_ulong *args)
88 {
89 target_ulong flags = args[0];
90 int shift = args[1];
91 uint64_t current_ram_size;
92 int rc;
93
94 if (spapr->resize_hpt == SPAPR_RESIZE_HPT_DISABLED) {
95 return H_AUTHORITY;
96 }
97
98 if (!spapr->htab_shift) {
99 /* Radix guest, no HPT */
100 return H_NOT_AVAILABLE;
101 }
102
103 trace_spapr_h_resize_hpt_prepare(flags, shift);
104
105 if (flags != 0) {
106 return H_PARAMETER;
107 }
108
109 if (shift && ((shift < 18) || (shift > 46))) {
110 return H_PARAMETER;
111 }
112
113 current_ram_size = MACHINE(spapr)->ram_size + get_plugged_memory_size();
114
115 /* We only allow the guest to allocate an HPT one order above what
116 * we'd normally give them (to stop a small guest claiming a huge
117 * chunk of resources in the HPT */
118 if (shift > (spapr_hpt_shift_for_ramsize(current_ram_size) + 1)) {
119 return H_RESOURCE;
120 }
121
122 rc = kvmppc_resize_hpt_prepare(cpu, flags, shift);
123 if (rc != -ENOSYS) {
124 return resize_hpt_convert_rc(rc);
125 }
126
127 if (kvm_enabled()) {
128 return H_HARDWARE;
129 } else if (tcg_enabled()) {
130 return vhyp_mmu_resize_hpt_prepare(cpu, spapr, shift);
131 } else {
132 g_assert_not_reached();
133 }
134 }
135
136 static void do_push_sregs_to_kvm_pr(CPUState *cs, run_on_cpu_data data)
137 {
138 int ret;
139
140 cpu_synchronize_state(cs);
141
142 ret = kvmppc_put_books_sregs(POWERPC_CPU(cs));
143 if (ret < 0) {
144 error_report("failed to push sregs to KVM: %s", strerror(-ret));
145 exit(1);
146 }
147 }
148
149 void push_sregs_to_kvm_pr(SpaprMachineState *spapr)
150 {
151 CPUState *cs;
152
153 /*
154 * This is a hack for the benefit of KVM PR - it abuses the SDR1
155 * slot in kvm_sregs to communicate the userspace address of the
156 * HPT
157 */
158 if (!kvm_enabled() || !spapr->htab) {
159 return;
160 }
161
162 CPU_FOREACH(cs) {
163 run_on_cpu(cs, do_push_sregs_to_kvm_pr, RUN_ON_CPU_NULL);
164 }
165 }
166
167 static target_ulong h_resize_hpt_commit(PowerPCCPU *cpu,
168 SpaprMachineState *spapr,
169 target_ulong opcode,
170 target_ulong *args)
171 {
172 target_ulong flags = args[0];
173 target_ulong shift = args[1];
174 int rc;
175
176 if (spapr->resize_hpt == SPAPR_RESIZE_HPT_DISABLED) {
177 return H_AUTHORITY;
178 }
179
180 if (!spapr->htab_shift) {
181 /* Radix guest, no HPT */
182 return H_NOT_AVAILABLE;
183 }
184
185 trace_spapr_h_resize_hpt_commit(flags, shift);
186
187 rc = kvmppc_resize_hpt_commit(cpu, flags, shift);
188 if (rc != -ENOSYS) {
189 rc = resize_hpt_convert_rc(rc);
190 if (rc == H_SUCCESS) {
191 /* Need to set the new htab_shift in the machine state */
192 spapr->htab_shift = shift;
193 }
194 return rc;
195 }
196
197 if (kvm_enabled()) {
198 return H_HARDWARE;
199 } else if (tcg_enabled()) {
200 return vhyp_mmu_resize_hpt_commit(cpu, spapr, flags, shift);
201 } else {
202 g_assert_not_reached();
203 }
204 }
205
206
207
208 static target_ulong h_set_sprg0(PowerPCCPU *cpu, SpaprMachineState *spapr,
209 target_ulong opcode, target_ulong *args)
210 {
211 cpu_synchronize_state(CPU(cpu));
212 cpu->env.spr[SPR_SPRG0] = args[0];
213
214 return H_SUCCESS;
215 }
216
217 static target_ulong h_set_dabr(PowerPCCPU *cpu, SpaprMachineState *spapr,
218 target_ulong opcode, target_ulong *args)
219 {
220 if (!ppc_has_spr(cpu, SPR_DABR)) {
221 return H_HARDWARE; /* DABR register not available */
222 }
223 cpu_synchronize_state(CPU(cpu));
224
225 if (ppc_has_spr(cpu, SPR_DABRX)) {
226 cpu->env.spr[SPR_DABRX] = 0x3; /* Use Problem and Privileged state */
227 } else if (!(args[0] & 0x4)) { /* Breakpoint Translation set? */
228 return H_RESERVED_DABR;
229 }
230
231 cpu->env.spr[SPR_DABR] = args[0];
232 return H_SUCCESS;
233 }
234
235 static target_ulong h_set_xdabr(PowerPCCPU *cpu, SpaprMachineState *spapr,
236 target_ulong opcode, target_ulong *args)
237 {
238 target_ulong dabrx = args[1];
239
240 if (!ppc_has_spr(cpu, SPR_DABR) || !ppc_has_spr(cpu, SPR_DABRX)) {
241 return H_HARDWARE;
242 }
243
244 if ((dabrx & ~0xfULL) != 0 || (dabrx & H_DABRX_HYPERVISOR) != 0
245 || (dabrx & (H_DABRX_KERNEL | H_DABRX_USER)) == 0) {
246 return H_PARAMETER;
247 }
248
249 cpu_synchronize_state(CPU(cpu));
250 cpu->env.spr[SPR_DABRX] = dabrx;
251 cpu->env.spr[SPR_DABR] = args[0];
252
253 return H_SUCCESS;
254 }
255
256 static target_ulong h_page_init(PowerPCCPU *cpu, SpaprMachineState *spapr,
257 target_ulong opcode, target_ulong *args)
258 {
259 target_ulong flags = args[0];
260 hwaddr dst = args[1];
261 hwaddr src = args[2];
262 hwaddr len = TARGET_PAGE_SIZE;
263 uint8_t *pdst, *psrc;
264 target_long ret = H_SUCCESS;
265
266 if (flags & ~(H_ICACHE_SYNCHRONIZE | H_ICACHE_INVALIDATE
267 | H_COPY_PAGE | H_ZERO_PAGE)) {
268 qemu_log_mask(LOG_UNIMP, "h_page_init: Bad flags (" TARGET_FMT_lx "\n",
269 flags);
270 return H_PARAMETER;
271 }
272
273 /* Map-in destination */
274 if (!is_ram_address(spapr, dst) || (dst & ~TARGET_PAGE_MASK) != 0) {
275 return H_PARAMETER;
276 }
277 pdst = physical_memory_map(dst, &len, true);
278 if (!pdst || len != TARGET_PAGE_SIZE) {
279 return H_PARAMETER;
280 }
281
282 if (flags & H_COPY_PAGE) {
283 /* Map-in source, copy to destination, and unmap source again */
284 if (!is_ram_address(spapr, src) || (src & ~TARGET_PAGE_MASK) != 0) {
285 ret = H_PARAMETER;
286 goto unmap_out;
287 }
288 psrc = physical_memory_map(src, &len, false);
289 if (!psrc || len != TARGET_PAGE_SIZE) {
290 ret = H_PARAMETER;
291 goto unmap_out;
292 }
293 memcpy(pdst, psrc, len);
294 physical_memory_unmap(psrc, len, 0, len);
295 } else if (flags & H_ZERO_PAGE) {
296 memset(pdst, 0, len); /* Just clear the destination page */
297 }
298
299 if (kvm_enabled() && (flags & H_ICACHE_SYNCHRONIZE) != 0) {
300 kvmppc_dcbst_range(cpu, pdst, len);
301 }
302 if (flags & (H_ICACHE_SYNCHRONIZE | H_ICACHE_INVALIDATE)) {
303 if (kvm_enabled()) {
304 kvmppc_icbi_range(cpu, pdst, len);
305 } else if (tcg_enabled()) {
306 tb_invalidate_phys_range(CPU(cpu), dst, dst + len - 1);
307 } else {
308 g_assert_not_reached();
309 }
310 }
311
312 unmap_out:
313 physical_memory_unmap(pdst, TARGET_PAGE_SIZE, 1, len);
314 return ret;
315 }
316
317 #define FLAGS_REGISTER_VPA 0x0000200000000000ULL
318 #define FLAGS_REGISTER_DTL 0x0000400000000000ULL
319 #define FLAGS_REGISTER_SLBSHADOW 0x0000600000000000ULL
320 #define FLAGS_DEREGISTER_VPA 0x0000a00000000000ULL
321 #define FLAGS_DEREGISTER_DTL 0x0000c00000000000ULL
322 #define FLAGS_DEREGISTER_SLBSHADOW 0x0000e00000000000ULL
323
324 static target_ulong register_vpa(PowerPCCPU *cpu, target_ulong vpa)
325 {
326 CPUState *cs = CPU(cpu);
327 CPUPPCState *env = &cpu->env;
328 SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
329 uint16_t size;
330 uint8_t tmp;
331
332 if (vpa == 0) {
333 hcall_dprintf("Can't cope with registering a VPA at logical 0\n");
334 return H_HARDWARE;
335 }
336
337 if (vpa % env->dcache_line_size) {
338 return H_PARAMETER;
339 }
340 /* FIXME: bounds check the address */
341
342 size = lduw_be_phys(cs->as, vpa + 0x4);
343
344 if (size < VPA_MIN_SIZE) {
345 return H_PARAMETER;
346 }
347
348 /* VPA is not allowed to cross a page boundary */
349 if ((vpa / 4096) != ((vpa + size - 1) / 4096)) {
350 return H_PARAMETER;
351 }
352
353 spapr_cpu->vpa_addr = vpa;
354
355 tmp = ldub_phys(cs->as, spapr_cpu->vpa_addr + VPA_SHARED_PROC_OFFSET);
356 tmp |= VPA_SHARED_PROC_VAL;
357 stb_phys(cs->as, spapr_cpu->vpa_addr + VPA_SHARED_PROC_OFFSET, tmp);
358
359 return H_SUCCESS;
360 }
361
362 static target_ulong deregister_vpa(PowerPCCPU *cpu, target_ulong vpa)
363 {
364 SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
365
366 if (spapr_cpu->slb_shadow_addr) {
367 return H_RESOURCE;
368 }
369
370 if (spapr_cpu->dtl_addr) {
371 return H_RESOURCE;
372 }
373
374 spapr_cpu->vpa_addr = 0;
375 return H_SUCCESS;
376 }
377
378 static target_ulong register_slb_shadow(PowerPCCPU *cpu, target_ulong addr)
379 {
380 SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
381 uint32_t size;
382
383 if (addr == 0) {
384 hcall_dprintf("Can't cope with SLB shadow at logical 0\n");
385 return H_HARDWARE;
386 }
387
388 size = ldl_be_phys(CPU(cpu)->as, addr + 0x4);
389 if (size < 0x8) {
390 return H_PARAMETER;
391 }
392
393 if ((addr / 4096) != ((addr + size - 1) / 4096)) {
394 return H_PARAMETER;
395 }
396
397 if (!spapr_cpu->vpa_addr) {
398 return H_RESOURCE;
399 }
400
401 spapr_cpu->slb_shadow_addr = addr;
402 spapr_cpu->slb_shadow_size = size;
403
404 return H_SUCCESS;
405 }
406
407 static target_ulong deregister_slb_shadow(PowerPCCPU *cpu, target_ulong addr)
408 {
409 SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
410
411 spapr_cpu->slb_shadow_addr = 0;
412 spapr_cpu->slb_shadow_size = 0;
413 return H_SUCCESS;
414 }
415
416 static target_ulong register_dtl(PowerPCCPU *cpu, target_ulong addr)
417 {
418 SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
419 uint32_t size;
420
421 if (addr == 0) {
422 hcall_dprintf("Can't cope with DTL at logical 0\n");
423 return H_HARDWARE;
424 }
425
426 size = ldl_be_phys(CPU(cpu)->as, addr + 0x4);
427
428 if (size < 48) {
429 return H_PARAMETER;
430 }
431
432 if (!spapr_cpu->vpa_addr) {
433 return H_RESOURCE;
434 }
435
436 spapr_cpu->dtl_addr = addr;
437 spapr_cpu->dtl_size = size;
438
439 return H_SUCCESS;
440 }
441
442 static target_ulong deregister_dtl(PowerPCCPU *cpu, target_ulong addr)
443 {
444 SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
445
446 spapr_cpu->dtl_addr = 0;
447 spapr_cpu->dtl_size = 0;
448
449 return H_SUCCESS;
450 }
451
452 static target_ulong h_register_vpa(PowerPCCPU *cpu, SpaprMachineState *spapr,
453 target_ulong opcode, target_ulong *args)
454 {
455 target_ulong flags = args[0];
456 target_ulong procno = args[1];
457 target_ulong vpa = args[2];
458 target_ulong ret = H_PARAMETER;
459 PowerPCCPU *tcpu;
460
461 tcpu = spapr_find_cpu(procno);
462 if (!tcpu) {
463 return H_PARAMETER;
464 }
465
466 switch (flags) {
467 case FLAGS_REGISTER_VPA:
468 ret = register_vpa(tcpu, vpa);
469 break;
470
471 case FLAGS_DEREGISTER_VPA:
472 ret = deregister_vpa(tcpu, vpa);
473 break;
474
475 case FLAGS_REGISTER_SLBSHADOW:
476 ret = register_slb_shadow(tcpu, vpa);
477 break;
478
479 case FLAGS_DEREGISTER_SLBSHADOW:
480 ret = deregister_slb_shadow(tcpu, vpa);
481 break;
482
483 case FLAGS_REGISTER_DTL:
484 ret = register_dtl(tcpu, vpa);
485 break;
486
487 case FLAGS_DEREGISTER_DTL:
488 ret = deregister_dtl(tcpu, vpa);
489 break;
490 }
491
492 return ret;
493 }
494
495 static target_ulong h_cede(PowerPCCPU *cpu, SpaprMachineState *spapr,
496 target_ulong opcode, target_ulong *args)
497 {
498 CPUPPCState *env = &cpu->env;
499 CPUState *cs = CPU(cpu);
500 SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
501
502 env->msr |= (1ULL << MSR_EE);
503 hreg_compute_hflags(env);
504 ppc_maybe_interrupt(env);
505
506 if (spapr_cpu->prod) {
507 spapr_cpu->prod = false;
508 return H_SUCCESS;
509 }
510
511 if (!cpu_has_work(cs)) {
512 cs->halted = 1;
513 cs->exception_index = EXCP_HLT;
514 ppc_maybe_interrupt(env);
515 cpu_exit(cs);
516 }
517
518 return H_SUCCESS;
519 }
520
521 /*
522 * Confer to self, aka join. Cede could use the same pattern as well, if
523 * EXCP_HLT can be changed to ECXP_HALTED.
524 */
525 static target_ulong h_confer_self(PowerPCCPU *cpu)
526 {
527 CPUState *cs = CPU(cpu);
528 SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
529
530 if (spapr_cpu->prod) {
531 spapr_cpu->prod = false;
532 return H_SUCCESS;
533 }
534 cs->halted = 1;
535 cs->exception_index = EXCP_HALTED;
536 ppc_maybe_interrupt(&cpu->env);
537 cpu_exit(cs);
538
539 return H_SUCCESS;
540 }
541
542 static target_ulong h_join(PowerPCCPU *cpu, SpaprMachineState *spapr,
543 target_ulong opcode, target_ulong *args)
544 {
545 CPUPPCState *env = &cpu->env;
546 CPUState *cs;
547 bool last_unjoined = true;
548
549 if (env->msr & (1ULL << MSR_EE)) {
550 return H_BAD_MODE;
551 }
552
553 /*
554 * Must not join the last CPU running. Interestingly, no such restriction
555 * for H_CONFER-to-self, but that is probably not intended to be used
556 * when H_JOIN is available.
557 */
558 CPU_FOREACH(cs) {
559 PowerPCCPU *c = POWERPC_CPU(cs);
560 CPUPPCState *e = &c->env;
561 if (c == cpu) {
562 continue;
563 }
564
565 /* Don't have a way to indicate joined, so use halted && MSR[EE]=0 */
566 if (!cs->halted || (e->msr & (1ULL << MSR_EE))) {
567 last_unjoined = false;
568 break;
569 }
570 }
571 if (last_unjoined) {
572 return H_CONTINUE;
573 }
574
575 return h_confer_self(cpu);
576 }
577
578 static target_ulong h_confer(PowerPCCPU *cpu, SpaprMachineState *spapr,
579 target_ulong opcode, target_ulong *args)
580 {
581 target_long target = args[0];
582 uint32_t dispatch = args[1];
583 CPUState *cs = CPU(cpu);
584 SpaprCpuState *spapr_cpu;
585
586 assert(tcg_enabled()); /* KVM will have handled this */
587
588 /*
589 * -1 means confer to all other CPUs without dispatch counter check,
590 * otherwise it's a targeted confer.
591 */
592 if (target != -1) {
593 PowerPCCPU *target_cpu = spapr_find_cpu(target);
594 uint32_t target_dispatch;
595
596 if (!target_cpu) {
597 return H_PARAMETER;
598 }
599
600 /*
601 * target == self is a special case, we wait until prodded, without
602 * dispatch counter check.
603 */
604 if (cpu == target_cpu) {
605 return h_confer_self(cpu);
606 }
607
608 spapr_cpu = spapr_cpu_state(target_cpu);
609 if (!spapr_cpu->vpa_addr || ((dispatch & 1) == 0)) {
610 return H_SUCCESS;
611 }
612
613 target_dispatch = ldl_be_phys(cs->as,
614 spapr_cpu->vpa_addr + VPA_DISPATCH_COUNTER);
615 if (target_dispatch != dispatch) {
616 return H_SUCCESS;
617 }
618
619 /*
620 * The targeted confer does not do anything special beyond yielding
621 * the current vCPU, but even this should be better than nothing.
622 * At least for single-threaded tcg, it gives the target a chance to
623 * run before we run again. Multi-threaded tcg does not really do
624 * anything with EXCP_YIELD yet.
625 */
626 }
627
628 cs->exception_index = EXCP_YIELD;
629 cpu_exit(cs);
630
631 return H_SUCCESS;
632 }
633
634 static target_ulong h_prod(PowerPCCPU *cpu, SpaprMachineState *spapr,
635 target_ulong opcode, target_ulong *args)
636 {
637 target_long target = args[0];
638 PowerPCCPU *tcpu;
639 CPUState *cs;
640 SpaprCpuState *spapr_cpu;
641
642 tcpu = spapr_find_cpu(target);
643 cs = CPU(tcpu);
644 if (!cs) {
645 return H_PARAMETER;
646 }
647
648 spapr_cpu = spapr_cpu_state(tcpu);
649 spapr_cpu->prod = true;
650 cs->halted = 0;
651 ppc_maybe_interrupt(&cpu->env);
652 qemu_cpu_kick(cs);
653
654 return H_SUCCESS;
655 }
656
657 static target_ulong h_rtas(PowerPCCPU *cpu, SpaprMachineState *spapr,
658 target_ulong opcode, target_ulong *args)
659 {
660 target_ulong rtas_r3 = args[0];
661 uint32_t token = rtas_ld(rtas_r3, 0);
662 uint32_t nargs = rtas_ld(rtas_r3, 1);
663 uint32_t nret = rtas_ld(rtas_r3, 2);
664
665 return spapr_rtas_call(cpu, spapr, token, nargs, rtas_r3 + 12,
666 nret, rtas_r3 + 12 + 4*nargs);
667 }
668
669 static target_ulong h_logical_load(PowerPCCPU *cpu, SpaprMachineState *spapr,
670 target_ulong opcode, target_ulong *args)
671 {
672 CPUState *cs = CPU(cpu);
673 target_ulong size = args[0];
674 target_ulong addr = args[1];
675
676 switch (size) {
677 case 1:
678 args[0] = ldub_phys(cs->as, addr);
679 return H_SUCCESS;
680 case 2:
681 args[0] = lduw_phys(cs->as, addr);
682 return H_SUCCESS;
683 case 4:
684 args[0] = ldl_phys(cs->as, addr);
685 return H_SUCCESS;
686 case 8:
687 args[0] = ldq_phys(cs->as, addr);
688 return H_SUCCESS;
689 }
690 return H_PARAMETER;
691 }
692
693 static target_ulong h_logical_store(PowerPCCPU *cpu, SpaprMachineState *spapr,
694 target_ulong opcode, target_ulong *args)
695 {
696 CPUState *cs = CPU(cpu);
697
698 target_ulong size = args[0];
699 target_ulong addr = args[1];
700 target_ulong val = args[2];
701
702 switch (size) {
703 case 1:
704 stb_phys(cs->as, addr, val);
705 return H_SUCCESS;
706 case 2:
707 stw_phys(cs->as, addr, val);
708 return H_SUCCESS;
709 case 4:
710 stl_phys(cs->as, addr, val);
711 return H_SUCCESS;
712 case 8:
713 stq_phys(cs->as, addr, val);
714 return H_SUCCESS;
715 }
716 return H_PARAMETER;
717 }
718
719 static target_ulong h_logical_memop(PowerPCCPU *cpu, SpaprMachineState *spapr,
720 target_ulong opcode, target_ulong *args)
721 {
722 CPUState *cs = CPU(cpu);
723
724 target_ulong dst = args[0]; /* Destination address */
725 target_ulong src = args[1]; /* Source address */
726 target_ulong esize = args[2]; /* Element size (0=1,1=2,2=4,3=8) */
727 target_ulong count = args[3]; /* Element count */
728 target_ulong op = args[4]; /* 0 = copy, 1 = invert */
729 uint64_t tmp;
730 unsigned int mask = (1 << esize) - 1;
731 int step = 1 << esize;
732
733 if (count > 0x80000000) {
734 return H_PARAMETER;
735 }
736
737 if ((dst & mask) || (src & mask) || (op > 1)) {
738 return H_PARAMETER;
739 }
740
741 if (dst >= src && dst < (src + (count << esize))) {
742 dst = dst + ((count - 1) << esize);
743 src = src + ((count - 1) << esize);
744 step = -step;
745 }
746
747 while (count--) {
748 switch (esize) {
749 case 0:
750 tmp = ldub_phys(cs->as, src);
751 break;
752 case 1:
753 tmp = lduw_phys(cs->as, src);
754 break;
755 case 2:
756 tmp = ldl_phys(cs->as, src);
757 break;
758 case 3:
759 tmp = ldq_phys(cs->as, src);
760 break;
761 default:
762 return H_PARAMETER;
763 }
764 if (op == 1) {
765 tmp = ~tmp;
766 }
767 switch (esize) {
768 case 0:
769 stb_phys(cs->as, dst, tmp);
770 break;
771 case 1:
772 stw_phys(cs->as, dst, tmp);
773 break;
774 case 2:
775 stl_phys(cs->as, dst, tmp);
776 break;
777 case 3:
778 stq_phys(cs->as, dst, tmp);
779 break;
780 }
781 dst = dst + step;
782 src = src + step;
783 }
784
785 return H_SUCCESS;
786 }
787
788 static target_ulong h_logical_icbi(PowerPCCPU *cpu, SpaprMachineState *spapr,
789 target_ulong opcode, target_ulong *args)
790 {
791 /* Nothing to do on emulation, KVM will trap this in the kernel */
792 return H_SUCCESS;
793 }
794
795 static target_ulong h_logical_dcbf(PowerPCCPU *cpu, SpaprMachineState *spapr,
796 target_ulong opcode, target_ulong *args)
797 {
798 /* Nothing to do on emulation, KVM will trap this in the kernel */
799 return H_SUCCESS;
800 }
801
802 static target_ulong h_set_mode_resource_set_ciabr(PowerPCCPU *cpu,
803 SpaprMachineState *spapr,
804 target_ulong mflags,
805 target_ulong value1,
806 target_ulong value2)
807 {
808 CPUPPCState *env = &cpu->env;
809
810 assert(tcg_enabled()); /* KVM will have handled this */
811
812 if (mflags) {
813 return H_UNSUPPORTED_FLAG;
814 }
815 if (value2) {
816 return H_P4;
817 }
818 if ((value1 & PPC_BITMASK(62, 63)) == 0x3) {
819 return H_P3;
820 }
821
822 ppc_store_ciabr(env, value1);
823
824 return H_SUCCESS;
825 }
826
827 static target_ulong h_set_mode_resource_set_dawr(PowerPCCPU *cpu,
828 SpaprMachineState *spapr,
829 target_ulong mflags,
830 target_ulong resource,
831 target_ulong value1,
832 target_ulong value2)
833 {
834 CPUPPCState *env = &cpu->env;
835
836 assert(tcg_enabled()); /* KVM will have handled this */
837
838 if (mflags) {
839 return H_UNSUPPORTED_FLAG;
840 }
841 if (value2 & PPC_BIT(61)) {
842 return H_P4;
843 }
844
845 if (resource == H_SET_MODE_RESOURCE_SET_DAWR0) {
846 ppc_store_dawr0(env, value1);
847 ppc_store_dawrx0(env, value2);
848 } else if (resource == H_SET_MODE_RESOURCE_SET_DAWR1) {
849 ppc_store_dawr1(env, value1);
850 ppc_store_dawrx1(env, value2);
851 } else {
852 g_assert_not_reached();
853 }
854
855 return H_SUCCESS;
856 }
857
858 static target_ulong h_set_mode_resource_le(PowerPCCPU *cpu,
859 SpaprMachineState *spapr,
860 target_ulong mflags,
861 target_ulong value1,
862 target_ulong value2)
863 {
864 if (value1) {
865 return H_P3;
866 }
867 if (value2) {
868 return H_P4;
869 }
870
871 switch (mflags) {
872 case H_SET_MODE_ENDIAN_BIG:
873 spapr_set_all_lpcrs(0, LPCR_ILE);
874 spapr_pci_switch_vga(spapr, true);
875 return H_SUCCESS;
876
877 case H_SET_MODE_ENDIAN_LITTLE:
878 spapr_set_all_lpcrs(LPCR_ILE, LPCR_ILE);
879 spapr_pci_switch_vga(spapr, false);
880 return H_SUCCESS;
881 }
882
883 return H_UNSUPPORTED_FLAG;
884 }
885
886 static target_ulong h_set_mode_resource_addr_trans_mode(PowerPCCPU *cpu,
887 SpaprMachineState *spapr,
888 target_ulong mflags,
889 target_ulong value1,
890 target_ulong value2)
891 {
892 if (value1) {
893 return H_P3;
894 }
895
896 if (value2) {
897 return H_P4;
898 }
899
900 /*
901 * AIL-1 is not architected, and AIL-2 is not supported by QEMU spapr.
902 * It is supported for faithful emulation of bare metal systems, but for
903 * compatibility concerns we leave it out of the pseries machine.
904 */
905 if (mflags != 0 && mflags != 3) {
906 return H_UNSUPPORTED_FLAG;
907 }
908
909 if (mflags == 3) {
910 if (!spapr_get_cap(spapr, SPAPR_CAP_AIL_MODE_3)) {
911 return H_UNSUPPORTED_FLAG;
912 }
913 }
914
915 spapr_set_all_lpcrs(mflags << LPCR_AIL_SHIFT, LPCR_AIL);
916
917 return H_SUCCESS;
918 }
919
920 static target_ulong h_set_mode(PowerPCCPU *cpu, SpaprMachineState *spapr,
921 target_ulong opcode, target_ulong *args)
922 {
923 target_ulong resource = args[1];
924 target_ulong ret = H_P2;
925
926 switch (resource) {
927 case H_SET_MODE_RESOURCE_SET_CIABR:
928 ret = h_set_mode_resource_set_ciabr(cpu, spapr, args[0], args[2],
929 args[3]);
930 break;
931 case H_SET_MODE_RESOURCE_SET_DAWR0:
932 case H_SET_MODE_RESOURCE_SET_DAWR1:
933 ret = h_set_mode_resource_set_dawr(cpu, spapr, args[0], args[1],
934 args[2], args[3]);
935 break;
936 case H_SET_MODE_RESOURCE_LE:
937 ret = h_set_mode_resource_le(cpu, spapr, args[0], args[2], args[3]);
938 break;
939 case H_SET_MODE_RESOURCE_ADDR_TRANS_MODE:
940 ret = h_set_mode_resource_addr_trans_mode(cpu, spapr, args[0],
941 args[2], args[3]);
942 break;
943 }
944
945 return ret;
946 }
947
948 static target_ulong h_clean_slb(PowerPCCPU *cpu, SpaprMachineState *spapr,
949 target_ulong opcode, target_ulong *args)
950 {
951 qemu_log_mask(LOG_UNIMP, "Unimplemented SPAPR hcall 0x"TARGET_FMT_lx"%s\n",
952 opcode, " (H_CLEAN_SLB)");
953 return H_FUNCTION;
954 }
955
956 static target_ulong h_invalidate_pid(PowerPCCPU *cpu, SpaprMachineState *spapr,
957 target_ulong opcode, target_ulong *args)
958 {
959 qemu_log_mask(LOG_UNIMP, "Unimplemented SPAPR hcall 0x"TARGET_FMT_lx"%s\n",
960 opcode, " (H_INVALIDATE_PID)");
961 return H_FUNCTION;
962 }
963
964 static void spapr_check_setup_free_hpt(SpaprMachineState *spapr,
965 uint64_t patbe_old, uint64_t patbe_new)
966 {
967 /*
968 * We have 4 Options:
969 * HASH->HASH || RADIX->RADIX || NOTHING->RADIX : Do Nothing
970 * HASH->RADIX : Free HPT
971 * RADIX->HASH : Allocate HPT
972 * NOTHING->HASH : Allocate HPT
973 * Note: NOTHING implies the case where we said the guest could choose
974 * later and so assumed radix and now it's called H_REG_PROC_TBL
975 */
976
977 if ((patbe_old & PATE1_GR) == (patbe_new & PATE1_GR)) {
978 /* We assume RADIX, so this catches all the "Do Nothing" cases */
979 } else if (!(patbe_old & PATE1_GR)) {
980 /* HASH->RADIX : Free HPT */
981 spapr_free_hpt(spapr);
982 } else if (!(patbe_new & PATE1_GR)) {
983 /* RADIX->HASH || NOTHING->HASH : Allocate HPT */
984 spapr_setup_hpt(spapr);
985 }
986 }
987
988 #define FLAGS_MASK 0x01FULL
989 #define FLAG_MODIFY 0x10
990 #define FLAG_REGISTER 0x08
991 #define FLAG_RADIX 0x04
992 #define FLAG_HASH_PROC_TBL 0x02
993 #define FLAG_GTSE 0x01
994
995 static target_ulong h_register_process_table(PowerPCCPU *cpu,
996 SpaprMachineState *spapr,
997 target_ulong opcode,
998 target_ulong *args)
999 {
1000 target_ulong flags = args[0];
1001 target_ulong proc_tbl = args[1];
1002 target_ulong page_size = args[2];
1003 target_ulong table_size = args[3];
1004 target_ulong update_lpcr = 0;
1005 target_ulong table_byte_size;
1006 uint64_t cproc;
1007
1008 if (flags & ~FLAGS_MASK) { /* Check no reserved bits are set */
1009 return H_PARAMETER;
1010 }
1011 if (flags & FLAG_MODIFY) {
1012 if (flags & FLAG_REGISTER) {
1013 /* Check process table alignment */
1014 table_byte_size = 1ULL << (table_size + 12);
1015 if (proc_tbl & (table_byte_size - 1)) {
1016 qemu_log_mask(LOG_GUEST_ERROR,
1017 "%s: process table not properly aligned: proc_tbl 0x"
1018 TARGET_FMT_lx" proc_tbl_size 0x"TARGET_FMT_lx"\n",
1019 __func__, proc_tbl, table_byte_size);
1020 }
1021 if (flags & FLAG_RADIX) { /* Register new RADIX process table */
1022 if (proc_tbl & 0xfff || proc_tbl >> 60) {
1023 return H_P2;
1024 } else if (page_size) {
1025 return H_P3;
1026 } else if (table_size > 24) {
1027 return H_P4;
1028 }
1029 cproc = PATE1_GR | proc_tbl | table_size;
1030 } else { /* Register new HPT process table */
1031 if (flags & FLAG_HASH_PROC_TBL) { /* Hash with Segment Tables */
1032 /* TODO - Not Supported */
1033 /* Technically caused by flag bits => H_PARAMETER */
1034 return H_PARAMETER;
1035 } else { /* Hash with SLB */
1036 if (proc_tbl >> 38) {
1037 return H_P2;
1038 } else if (page_size & ~0x7) {
1039 return H_P3;
1040 } else if (table_size > 24) {
1041 return H_P4;
1042 }
1043 }
1044 cproc = (proc_tbl << 25) | page_size << 5 | table_size;
1045 }
1046
1047 } else { /* Deregister current process table */
1048 /*
1049 * Set to benign value: (current GR) | 0. This allows
1050 * deregistration in KVM to succeed even if the radix bit
1051 * in flags doesn't match the radix bit in the old PATE.
1052 */
1053 cproc = spapr->patb_entry & PATE1_GR;
1054 }
1055 } else { /* Maintain current registration */
1056 if (!(flags & FLAG_RADIX) != !(spapr->patb_entry & PATE1_GR)) {
1057 /* Technically caused by flag bits => H_PARAMETER */
1058 return H_PARAMETER; /* Existing Process Table Mismatch */
1059 }
1060 cproc = spapr->patb_entry;
1061 }
1062
1063 /* Check if we need to setup OR free the hpt */
1064 spapr_check_setup_free_hpt(spapr, spapr->patb_entry, cproc);
1065
1066 spapr->patb_entry = cproc; /* Save new process table */
1067
1068 /* Update the UPRT, HR and GTSE bits in the LPCR for all cpus */
1069 if (flags & FLAG_RADIX) /* Radix must use process tables, also set HR */
1070 update_lpcr |= (LPCR_UPRT | LPCR_HR);
1071 else if (flags & FLAG_HASH_PROC_TBL) /* Hash with process tables */
1072 update_lpcr |= LPCR_UPRT;
1073 if (flags & FLAG_GTSE) /* Guest translation shootdown enable */
1074 update_lpcr |= LPCR_GTSE;
1075
1076 spapr_set_all_lpcrs(update_lpcr, LPCR_UPRT | LPCR_HR | LPCR_GTSE);
1077
1078 if (kvm_enabled()) {
1079 return kvmppc_configure_v3_mmu(cpu, flags & FLAG_RADIX,
1080 flags & FLAG_GTSE, cproc);
1081 }
1082 return H_SUCCESS;
1083 }
1084
1085 #define H_SIGNAL_SYS_RESET_ALL -1
1086 #define H_SIGNAL_SYS_RESET_ALLBUTSELF -2
1087
1088 static target_ulong h_signal_sys_reset(PowerPCCPU *cpu,
1089 SpaprMachineState *spapr,
1090 target_ulong opcode, target_ulong *args)
1091 {
1092 target_long target = args[0];
1093 CPUState *cs;
1094
1095 if (target < 0) {
1096 /* Broadcast */
1097 if (target < H_SIGNAL_SYS_RESET_ALLBUTSELF) {
1098 return H_PARAMETER;
1099 }
1100
1101 CPU_FOREACH(cs) {
1102 PowerPCCPU *c = POWERPC_CPU(cs);
1103
1104 if (target == H_SIGNAL_SYS_RESET_ALLBUTSELF) {
1105 if (c == cpu) {
1106 continue;
1107 }
1108 }
1109
1110 /* Skip quiesced CPUs - they are in RTAS stopped state and
1111 * should not be reset. This prevents kdump hangs when CPUs
1112 * are hotplugged but not yet started by the guest.
1113 */
1114 if (c->env.quiesced) {
1115 continue;
1116 }
1117
1118 run_on_cpu(cs, spapr_do_system_reset_on_cpu, RUN_ON_CPU_NULL);
1119 }
1120 return H_SUCCESS;
1121
1122 } else {
1123 /* Unicast */
1124 cs = CPU(spapr_find_cpu(target));
1125 if (cs) {
1126 run_on_cpu(cs, spapr_do_system_reset_on_cpu, RUN_ON_CPU_NULL);
1127 return H_SUCCESS;
1128 }
1129 return H_PARAMETER;
1130 }
1131 }
1132
1133 /* Returns either a logical PVR or zero if none was found */
1134 static uint32_t cas_check_pvr(PowerPCCPU *cpu, uint32_t max_compat,
1135 target_ulong *addr, bool *raw_mode_supported)
1136 {
1137 bool explicit_match = false; /* Matched the CPU's real PVR */
1138 uint32_t best_compat = 0;
1139 uint32_t compat_host_pvr = 0;
1140 int i;
1141
1142 /*
1143 * We scan the supplied table of PVRs looking for two things
1144 * 1. Is our real CPU PVR in the list?
1145 * 2. What's the "best" listed logical PVR
1146 */
1147 for (i = 0; i < 512; ++i) {
1148 uint32_t pvr, pvr_mask;
1149
1150 pvr_mask = ldl_be_phys(&address_space_memory, *addr);
1151 pvr = ldl_be_phys(&address_space_memory, *addr + 4);
1152 *addr += 8;
1153
1154 if (~pvr_mask & pvr) {
1155 break; /* Terminator record */
1156 }
1157
1158 if ((cpu->env.spr[SPR_PVR] & pvr_mask) == (pvr & pvr_mask)) {
1159 explicit_match = true;
1160 } else {
1161 if (ppc_check_compat(cpu, pvr, best_compat, max_compat)) {
1162 best_compat = pvr;
1163 }
1164 }
1165 }
1166
1167 if (explicit_match && kvm_enabled()) {
1168 compat_host_pvr = kvm_ppc_host_compat_pvr();
1169 /*
1170 * If the host is booted in a compatibility mode, do not try booting in
1171 * the raw mode as it may allow KVM guests to boot with a higher CPU
1172 * version compared to what host was booted with; which should not be
1173 * allowed.
1174 */
1175 if (compat_host_pvr) {
1176 explicit_match = false;
1177 }
1178 }
1179
1180 *raw_mode_supported = explicit_match;
1181
1182 /* Parsing finished */
1183 trace_spapr_cas_pvr(cpu->compat_pvr, explicit_match, best_compat);
1184
1185 return best_compat;
1186 }
1187
1188 static
1189 target_ulong do_client_architecture_support(PowerPCCPU *cpu,
1190 SpaprMachineState *spapr,
1191 target_ulong vec,
1192 target_ulong fdt_bufsize)
1193 {
1194 target_ulong ov_table; /* Working address in data buffer */
1195 uint32_t cas_pvr;
1196 SpaprOptionVector *ov1_guest, *ov5_guest;
1197 bool guest_radix;
1198 bool raw_mode_supported = false;
1199 bool guest_xive;
1200 CPUState *cs;
1201 void *fdt;
1202 uint32_t max_compat = spapr->max_compat_pvr;
1203
1204 /* CAS is supposed to be called early when only the boot vCPU is active. */
1205 CPU_FOREACH(cs) {
1206 if (cs == CPU(cpu)) {
1207 continue;
1208 }
1209 if (!cs->halted) {
1210 warn_report("guest has multiple active vCPUs at CAS, which is not allowed");
1211 return H_MULTI_THREADS_ACTIVE;
1212 }
1213 }
1214
1215 cas_pvr = cas_check_pvr(cpu, max_compat, &vec, &raw_mode_supported);
1216 if (!cas_pvr && (!raw_mode_supported || max_compat)) {
1217 /*
1218 * We couldn't find a suitable compatibility mode, and either
1219 * the guest doesn't support "raw" mode for this CPU, or "raw"
1220 * mode is disabled because a maximum compat mode is set.
1221 */
1222 error_report("Couldn't negotiate a suitable PVR during CAS");
1223 return H_HARDWARE;
1224 }
1225
1226 /* Update CPUs */
1227 if (cpu->compat_pvr != cas_pvr) {
1228 Error *local_err = NULL;
1229
1230 if (ppc_set_compat_all(cas_pvr, &local_err) < 0) {
1231 /* We fail to set compat mode (likely because running with KVM PR),
1232 * but maybe we can fallback to raw mode if the guest supports it.
1233 */
1234 if (!raw_mode_supported) {
1235 error_report_err(local_err);
1236 return H_HARDWARE;
1237 }
1238 error_free(local_err);
1239 }
1240 }
1241
1242 /* For the future use: here @ov_table points to the first option vector */
1243 ov_table = vec;
1244
1245 ov1_guest = spapr_ovec_parse_vector(ov_table, 1);
1246 if (!ov1_guest) {
1247 warn_report("guest didn't provide option vector 1");
1248 return H_PARAMETER;
1249 }
1250 ov5_guest = spapr_ovec_parse_vector(ov_table, 5);
1251 if (!ov5_guest) {
1252 spapr_ovec_cleanup(ov1_guest);
1253 warn_report("guest didn't provide option vector 5");
1254 return H_PARAMETER;
1255 }
1256 if (spapr_ovec_test(ov5_guest, OV5_MMU_BOTH)) {
1257 error_report("guest requested hash and radix MMU, which is invalid.");
1258 exit(EXIT_FAILURE);
1259 }
1260 if (spapr_ovec_test(ov5_guest, OV5_XIVE_BOTH)) {
1261 error_report("guest requested an invalid interrupt mode");
1262 exit(EXIT_FAILURE);
1263 }
1264
1265 guest_radix = spapr_ovec_test(ov5_guest, OV5_MMU_RADIX_300);
1266
1267 guest_xive = spapr_ovec_test(ov5_guest, OV5_XIVE_EXPLOIT);
1268
1269 /*
1270 * HPT resizing is a bit of a special case, because when enabled
1271 * we assume an HPT guest will support it until it says it
1272 * doesn't, instead of assuming it won't support it until it says
1273 * it does. Strictly speaking that approach could break for
1274 * guests which don't make a CAS call, but those are so old we
1275 * don't care about them. Without that assumption we'd have to
1276 * make at least a temporary allocation of an HPT sized for max
1277 * memory, which could be impossibly difficult under KVM HV if
1278 * maxram is large.
1279 */
1280 if (!guest_radix && !spapr_ovec_test(ov5_guest, OV5_HPT_RESIZE)) {
1281 int maxshift = spapr_hpt_shift_for_ramsize(MACHINE(spapr)->maxram_size);
1282
1283 if (spapr->resize_hpt == SPAPR_RESIZE_HPT_REQUIRED) {
1284 error_report(
1285 "h_client_architecture_support: Guest doesn't support HPT resizing, but resize-hpt=required");
1286 exit(1);
1287 }
1288
1289 if (spapr->htab_shift < maxshift) {
1290 /* Guest doesn't know about HPT resizing, so we
1291 * pre-emptively resize for the maximum permitted RAM. At
1292 * the point this is called, nothing should have been
1293 * entered into the existing HPT */
1294 spapr_reallocate_hpt(spapr, maxshift, &error_fatal);
1295 push_sregs_to_kvm_pr(spapr);
1296 }
1297 }
1298
1299 /* NOTE: there are actually a number of ov5 bits where input from the
1300 * guest is always zero, and the platform/QEMU enables them independently
1301 * of guest input. To model these properly we'd want some sort of mask,
1302 * but since they only currently apply to memory migration as defined
1303 * by LoPAPR 1.1, 14.5.4.8, which QEMU doesn't implement, we don't need
1304 * to worry about this for now.
1305 */
1306
1307 /* full range of negotiated ov5 capabilities */
1308 spapr_ovec_intersect(spapr->ov5_cas, spapr->ov5, ov5_guest);
1309 spapr_ovec_cleanup(ov5_guest);
1310
1311 spapr_check_mmu_mode(guest_radix);
1312
1313 spapr->cas_pre_isa3_guest = !spapr_ovec_test(ov1_guest, OV1_PPC_3_00);
1314 spapr_ovec_cleanup(ov1_guest);
1315
1316 /*
1317 * Check for NUMA affinity conditions now that we know which NUMA
1318 * affinity the guest will use.
1319 */
1320 spapr_numa_associativity_check(spapr);
1321
1322 /*
1323 * Ensure the guest asks for an interrupt mode we support;
1324 * otherwise terminate the boot.
1325 */
1326 if (guest_xive) {
1327 if (!spapr->irq->xive) {
1328 error_report(
1329 "Guest requested unavailable interrupt mode (XIVE), try the ic-mode=xive or ic-mode=dual machine property");
1330 exit(EXIT_FAILURE);
1331 }
1332 } else {
1333 if (!spapr->irq->xics) {
1334 error_report(
1335 "Guest requested unavailable interrupt mode (XICS), either don't set the ic-mode machine property or try ic-mode=xics or ic-mode=dual");
1336 exit(EXIT_FAILURE);
1337 }
1338 }
1339
1340 spapr_irq_update_active_intc(spapr);
1341
1342 /*
1343 * Process all pending hot-plug/unplug requests now. An updated full
1344 * rendered FDT will be returned to the guest.
1345 */
1346 spapr_drc_reset_all(spapr);
1347 spapr_clear_pending_hotplug_events(spapr);
1348
1349 /*
1350 * If spapr_machine_reset() did not set up a HPT but one is necessary
1351 * (because the guest isn't going to use radix) then set it up here.
1352 */
1353 if ((spapr->patb_entry & PATE1_GR) && !guest_radix) {
1354 /* legacy hash or new hash: */
1355 spapr_setup_hpt(spapr);
1356 }
1357
1358 fdt = spapr_build_fdt(spapr, spapr->vof != NULL, fdt_bufsize);
1359 g_free(spapr->fdt_blob);
1360 spapr->fdt_size = fdt_totalsize(fdt);
1361 spapr->fdt_initial_size = spapr->fdt_size;
1362 spapr->fdt_blob = fdt;
1363
1364 /*
1365 * Set the machine->fdt pointer again since we just freed
1366 * it above (by freeing spapr->fdt_blob). We set this
1367 * pointer to enable support for the 'dumpdtb' QMP/HMP
1368 * command.
1369 */
1370 MACHINE(spapr)->fdt = fdt;
1371
1372 return H_SUCCESS;
1373 }
1374
1375 static target_ulong h_client_architecture_support(PowerPCCPU *cpu,
1376 SpaprMachineState *spapr,
1377 target_ulong opcode,
1378 target_ulong *args)
1379 {
1380 target_ulong vec = ppc64_phys_to_real(args[0]);
1381 target_ulong fdt_buf = args[1];
1382 target_ulong fdt_bufsize = args[2];
1383 target_ulong ret;
1384 SpaprDeviceTreeUpdateHeader hdr = { .version_id = 1 };
1385
1386 if (fdt_bufsize < sizeof(hdr)) {
1387 error_report("SLOF provided insufficient CAS buffer "
1388 TARGET_FMT_lu " (min: %zu)", fdt_bufsize, sizeof(hdr));
1389 exit(EXIT_FAILURE);
1390 }
1391
1392 fdt_bufsize -= sizeof(hdr);
1393
1394 ret = do_client_architecture_support(cpu, spapr, vec, fdt_bufsize);
1395 if (ret == H_SUCCESS) {
1396 _FDT((fdt_pack(spapr->fdt_blob)));
1397 spapr->fdt_size = fdt_totalsize(spapr->fdt_blob);
1398 spapr->fdt_initial_size = spapr->fdt_size;
1399
1400 physical_memory_write(fdt_buf, &hdr, sizeof(hdr));
1401 physical_memory_write(fdt_buf + sizeof(hdr), spapr->fdt_blob,
1402 spapr->fdt_size);
1403 trace_spapr_cas_continue(spapr->fdt_size + sizeof(hdr));
1404 }
1405
1406 return ret;
1407 }
1408
1409 target_ulong spapr_vof_client_architecture_support(MachineState *ms,
1410 CPUState *cs,
1411 target_ulong ovec_addr)
1412 {
1413 SpaprMachineState *spapr = SPAPR_MACHINE(ms);
1414
1415 target_ulong ret = do_client_architecture_support(POWERPC_CPU(cs), spapr,
1416 ovec_addr, FDT_MAX_SIZE);
1417
1418 /*
1419 * This adds stdout and generates phandles for boottime and CAS FDTs.
1420 * It is alright to update the FDT here as do_client_architecture_support()
1421 * does not pack it.
1422 */
1423 spapr_vof_client_dt_finalize(spapr, spapr->fdt_blob);
1424
1425 return ret;
1426 }
1427
1428 static target_ulong h_get_cpu_characteristics(PowerPCCPU *cpu,
1429 SpaprMachineState *spapr,
1430 target_ulong opcode,
1431 target_ulong *args)
1432 {
1433 uint64_t characteristics = H_CPU_CHAR_HON_BRANCH_HINTS &
1434 ~H_CPU_CHAR_THR_RECONF_TRIG;
1435 uint64_t behaviour = H_CPU_BEHAV_FAVOUR_SECURITY;
1436 uint8_t safe_cache = spapr_get_cap(spapr, SPAPR_CAP_CFPC);
1437 uint8_t safe_bounds_check = spapr_get_cap(spapr, SPAPR_CAP_SBBC);
1438 uint8_t safe_indirect_branch = spapr_get_cap(spapr, SPAPR_CAP_IBS);
1439 uint8_t count_cache_flush_assist = spapr_get_cap(spapr,
1440 SPAPR_CAP_CCF_ASSIST);
1441
1442 switch (safe_cache) {
1443 case SPAPR_CAP_WORKAROUND:
1444 characteristics |= H_CPU_CHAR_L1D_FLUSH_ORI30;
1445 characteristics |= H_CPU_CHAR_L1D_FLUSH_TRIG2;
1446 characteristics |= H_CPU_CHAR_L1D_THREAD_PRIV;
1447 behaviour |= H_CPU_BEHAV_L1D_FLUSH_PR;
1448 break;
1449 case SPAPR_CAP_FIXED:
1450 behaviour |= H_CPU_BEHAV_NO_L1D_FLUSH_ENTRY;
1451 behaviour |= H_CPU_BEHAV_NO_L1D_FLUSH_UACCESS;
1452 break;
1453 default: /* broken */
1454 assert(safe_cache == SPAPR_CAP_BROKEN);
1455 behaviour |= H_CPU_BEHAV_L1D_FLUSH_PR;
1456 break;
1457 }
1458
1459 switch (safe_bounds_check) {
1460 case SPAPR_CAP_WORKAROUND:
1461 characteristics |= H_CPU_CHAR_SPEC_BAR_ORI31;
1462 behaviour |= H_CPU_BEHAV_BNDS_CHK_SPEC_BAR;
1463 break;
1464 case SPAPR_CAP_FIXED:
1465 break;
1466 default: /* broken */
1467 assert(safe_bounds_check == SPAPR_CAP_BROKEN);
1468 behaviour |= H_CPU_BEHAV_BNDS_CHK_SPEC_BAR;
1469 break;
1470 }
1471
1472 switch (safe_indirect_branch) {
1473 case SPAPR_CAP_FIXED_NA:
1474 break;
1475 case SPAPR_CAP_FIXED_CCD:
1476 characteristics |= H_CPU_CHAR_CACHE_COUNT_DIS;
1477 break;
1478 case SPAPR_CAP_FIXED_IBS:
1479 characteristics |= H_CPU_CHAR_BCCTRL_SERIALISED;
1480 break;
1481 case SPAPR_CAP_WORKAROUND:
1482 behaviour |= H_CPU_BEHAV_FLUSH_COUNT_CACHE;
1483 if (count_cache_flush_assist) {
1484 characteristics |= H_CPU_CHAR_BCCTR_FLUSH_ASSIST;
1485 }
1486 break;
1487 default: /* broken */
1488 assert(safe_indirect_branch == SPAPR_CAP_BROKEN);
1489 break;
1490 }
1491
1492 args[0] = characteristics;
1493 args[1] = behaviour;
1494 return H_SUCCESS;
1495 }
1496
1497 static target_ulong h_update_dt(PowerPCCPU *cpu, SpaprMachineState *spapr,
1498 target_ulong opcode, target_ulong *args)
1499 {
1500 target_ulong dt = ppc64_phys_to_real(args[0]);
1501 struct fdt_header hdr = { 0 };
1502 unsigned cb;
1503 void *fdt;
1504
1505 physical_memory_read(dt, &hdr, sizeof(hdr));
1506 cb = fdt32_to_cpu(hdr.totalsize);
1507
1508 /* Check that the fdt did not grow out of proportion */
1509 if (cb > spapr->fdt_initial_size * 2) {
1510 trace_spapr_update_dt_failed_size(spapr->fdt_initial_size, cb,
1511 fdt32_to_cpu(hdr.magic));
1512 return H_PARAMETER;
1513 }
1514
1515 fdt = g_malloc0(cb);
1516 physical_memory_read(dt, fdt, cb);
1517
1518 /* Check the fdt consistency */
1519 if (fdt_check_full(fdt, cb)) {
1520 trace_spapr_update_dt_failed_check(spapr->fdt_initial_size, cb,
1521 fdt32_to_cpu(hdr.magic));
1522 return H_PARAMETER;
1523 }
1524
1525 g_free(spapr->fdt_blob);
1526 spapr->fdt_size = cb;
1527 spapr->fdt_blob = fdt;
1528 trace_spapr_update_dt(cb);
1529
1530 return H_SUCCESS;
1531 }
1532
1533 static spapr_hcall_fn papr_hypercall_table[(MAX_HCALL_OPCODE / 4) + 1];
1534 static spapr_hcall_fn kvmppc_hypercall_table[KVMPPC_HCALL_MAX - KVMPPC_HCALL_BASE + 1];
1535 static spapr_hcall_fn svm_hypercall_table[(SVM_HCALL_MAX - SVM_HCALL_BASE) / 4 + 1];
1536
1537 void spapr_register_hypercall(target_ulong opcode, spapr_hcall_fn fn)
1538 {
1539 spapr_hcall_fn *slot;
1540
1541 if (opcode <= MAX_HCALL_OPCODE) {
1542 assert((opcode & 0x3) == 0);
1543
1544 slot = &papr_hypercall_table[opcode / 4];
1545 } else if (opcode >= SVM_HCALL_BASE && opcode <= SVM_HCALL_MAX) {
1546 /* we only have SVM-related hcall numbers assigned in multiples of 4 */
1547 assert((opcode & 0x3) == 0);
1548
1549 slot = &svm_hypercall_table[(opcode - SVM_HCALL_BASE) / 4];
1550 } else {
1551 assert((opcode >= KVMPPC_HCALL_BASE) && (opcode <= KVMPPC_HCALL_MAX));
1552
1553 slot = &kvmppc_hypercall_table[opcode - KVMPPC_HCALL_BASE];
1554 }
1555
1556 assert(!(*slot));
1557 *slot = fn;
1558 }
1559
1560 void spapr_unregister_hypercall(target_ulong opcode)
1561 {
1562 spapr_hcall_fn *slot;
1563
1564 if (opcode <= MAX_HCALL_OPCODE) {
1565 assert((opcode & 0x3) == 0);
1566
1567 slot = &papr_hypercall_table[opcode / 4];
1568 } else if (opcode >= SVM_HCALL_BASE && opcode <= SVM_HCALL_MAX) {
1569 /* we only have SVM-related hcall numbers assigned in multiples of 4 */
1570 assert((opcode & 0x3) == 0);
1571
1572 slot = &svm_hypercall_table[(opcode - SVM_HCALL_BASE) / 4];
1573 } else {
1574 assert((opcode >= KVMPPC_HCALL_BASE) && (opcode <= KVMPPC_HCALL_MAX));
1575
1576 slot = &kvmppc_hypercall_table[opcode - KVMPPC_HCALL_BASE];
1577 }
1578
1579 *slot = NULL;
1580 }
1581
1582 target_ulong spapr_hypercall(PowerPCCPU *cpu, target_ulong opcode,
1583 target_ulong *args)
1584 {
1585 SpaprMachineState *spapr = SPAPR_MACHINE(qdev_get_machine());
1586
1587 if ((opcode <= MAX_HCALL_OPCODE)
1588 && ((opcode & 0x3) == 0)) {
1589 spapr_hcall_fn fn = papr_hypercall_table[opcode / 4];
1590
1591 if (fn) {
1592 return fn(cpu, spapr, opcode, args);
1593 }
1594 } else if ((opcode >= SVM_HCALL_BASE) &&
1595 (opcode <= SVM_HCALL_MAX)) {
1596 spapr_hcall_fn fn = svm_hypercall_table[(opcode - SVM_HCALL_BASE) / 4];
1597
1598 if (fn) {
1599 return fn(cpu, spapr, opcode, args);
1600 }
1601 } else if ((opcode >= KVMPPC_HCALL_BASE) &&
1602 (opcode <= KVMPPC_HCALL_MAX)) {
1603 spapr_hcall_fn fn = kvmppc_hypercall_table[opcode - KVMPPC_HCALL_BASE];
1604
1605 if (fn) {
1606 return fn(cpu, spapr, opcode, args);
1607 }
1608 }
1609
1610 qemu_log_mask(LOG_UNIMP, "Unimplemented SPAPR hcall 0x" TARGET_FMT_lx "\n",
1611 opcode);
1612 return H_FUNCTION;
1613 }
1614
1615 #ifdef CONFIG_TCG
1616 static void hypercall_register_softmmu(void)
1617 {
1618 /* DO NOTHING */
1619 }
1620 #else
1621 static target_ulong h_softmmu(PowerPCCPU *cpu, SpaprMachineState *spapr,
1622 target_ulong opcode, target_ulong *args)
1623 {
1624 g_assert_not_reached();
1625 }
1626
1627 static void hypercall_register_softmmu(void)
1628 {
1629 /* hcall-pft */
1630 spapr_register_hypercall(H_ENTER, h_softmmu);
1631 spapr_register_hypercall(H_REMOVE, h_softmmu);
1632 spapr_register_hypercall(H_PROTECT, h_softmmu);
1633 spapr_register_hypercall(H_READ, h_softmmu);
1634
1635 /* hcall-bulk */
1636 spapr_register_hypercall(H_BULK_REMOVE, h_softmmu);
1637 }
1638 #endif
1639
1640 static void hypercall_register_types(void)
1641 {
1642 hypercall_register_softmmu();
1643
1644 /* hcall-hpt-resize */
1645 spapr_register_hypercall(H_RESIZE_HPT_PREPARE, h_resize_hpt_prepare);
1646 spapr_register_hypercall(H_RESIZE_HPT_COMMIT, h_resize_hpt_commit);
1647
1648 /* hcall-splpar */
1649 spapr_register_hypercall(H_REGISTER_VPA, h_register_vpa);
1650 spapr_register_hypercall(H_CEDE, h_cede);
1651 spapr_register_hypercall(H_CONFER, h_confer);
1652 spapr_register_hypercall(H_PROD, h_prod);
1653
1654 /* hcall-join */
1655 spapr_register_hypercall(H_JOIN, h_join);
1656
1657 spapr_register_hypercall(H_SIGNAL_SYS_RESET, h_signal_sys_reset);
1658
1659 /* processor register resource access h-calls */
1660 spapr_register_hypercall(H_SET_SPRG0, h_set_sprg0);
1661 spapr_register_hypercall(H_SET_DABR, h_set_dabr);
1662 spapr_register_hypercall(H_SET_XDABR, h_set_xdabr);
1663 spapr_register_hypercall(H_PAGE_INIT, h_page_init);
1664 spapr_register_hypercall(H_SET_MODE, h_set_mode);
1665
1666 /* In Memory Table MMU h-calls */
1667 spapr_register_hypercall(H_CLEAN_SLB, h_clean_slb);
1668 spapr_register_hypercall(H_INVALIDATE_PID, h_invalidate_pid);
1669 spapr_register_hypercall(H_REGISTER_PROC_TBL, h_register_process_table);
1670
1671 /* hcall-get-cpu-characteristics */
1672 spapr_register_hypercall(H_GET_CPU_CHARACTERISTICS,
1673 h_get_cpu_characteristics);
1674
1675 /* "debugger" hcalls (also used by SLOF). Note: We do -not- differentiate
1676 * here between the "CI" and the "CACHE" variants, they will use whatever
1677 * mapping attributes qemu is using. When using KVM, the kernel will
1678 * enforce the attributes more strongly
1679 */
1680 spapr_register_hypercall(H_LOGICAL_CI_LOAD, h_logical_load);
1681 spapr_register_hypercall(H_LOGICAL_CI_STORE, h_logical_store);
1682 spapr_register_hypercall(H_LOGICAL_CACHE_LOAD, h_logical_load);
1683 spapr_register_hypercall(H_LOGICAL_CACHE_STORE, h_logical_store);
1684 spapr_register_hypercall(H_LOGICAL_ICBI, h_logical_icbi);
1685 spapr_register_hypercall(H_LOGICAL_DCBF, h_logical_dcbf);
1686 spapr_register_hypercall(KVMPPC_H_LOGICAL_MEMOP, h_logical_memop);
1687
1688 /* qemu/KVM-PPC specific hcalls */
1689 spapr_register_hypercall(KVMPPC_H_RTAS, h_rtas);
1690
1691 /* ibm,client-architecture-support support */
1692 spapr_register_hypercall(KVMPPC_H_CAS, h_client_architecture_support);
1693
1694 spapr_register_hypercall(KVMPPC_H_UPDATE_DT, h_update_dt);
1695 }
1696
1697 type_init(hypercall_register_types)